Report No. 51341-TJ Republic of Tajikistan Poverty Assessment December 3, 2009 Human Development Sector Unit Central Asia Country Unit Europe and Central Asia Region Document of the World Bank Table of Contents Executive Summary ....................................................................................................................... 1 A. Poverty Dynamics............................................................................................................ 2 B. Poverty Profile ................................................................................................................. 4 C. Domestic Labor Market ................................................................................................... 7 D. Migration ......................................................................................................................... 8 E. Social Protection .............................................................................................................. 9 F. Education ....................................................................................................................... 10 G. Health ............................................................................................................................ 10 Chapter 1: Poverty Dynamics in Tajikistan .............................................................................. 12 A. Introduction ................................................................................................................... 12 B. Economic Growth and Trends in Consumption between 2003 and 2007 ..................... 12 C. The Expected Poverty Impact of the Financial Crisis ................................................... 18 D. Conclusions ................................................................................................................... 20 Chapter 2: Poverty Profile in Tajikistan ................................................................................... 22 A. Introduction ................................................................................................................... 22 B. Poverty and Inequality in Tajikistan .............................................................................. 22 C. Poverty Profile and Multivariate Analysis of Poverty ................................................... 25 D. Household Income Poverty............................................................................................ 29 E. Other Evidence of Living Standards.............................................................................. 31 F. Conclusion ..................................................................................................................... 34 Chapter 3: Tajikistan Domestic Labor Market ........................................................................ 35 A. Introduction ................................................................................................................... 35 B. General Labor Market Issues ......................................................................................... 35 C. Non-participation ........................................................................................................... 38 D. Unemployment .............................................................................................................. 39 E. Employment................................................................................................................... 41 F. Labor Incomes and the Working Poor ........................................................................... 43 G. Conclusions ................................................................................................................... 45 Chapter 4: Labor Migration, Remittances, and Welfare Implications for Tajikistan .......... 47 A. Introduction ................................................................................................................... 47 B. Internal and International Migration Flows in Tajikistan .............................................. 47 C. The Welfare Implications of Remittances for Households ............................................ 53 D. Conclusions ................................................................................................................... 55 Chapter 5: Social Protection in Tajikistan ................................................................................ 57 A. Introduction ................................................................................................................... 57 B. Coverage of Social Protection ....................................................................................... 58 C. Incidence of Benefits ..................................................................................................... 61 D. Adequacy and Poverty Impact of Social Protection (SP) Benefits ................................ 62 E. Toward a More Comprehensive and Effective Social Protection (SP) System ............. 65 F. Conclusions ................................................................................................................... 67 ii Chapter 6: Education in Tajikistan ­ Progress but Many More Challenges Ahead ............. 69 A. Introduction ................................................................................................................... 69 B. Public and Private Expenditure on Education ............................................................... 70 C. Enrollment, with a Poverty and Gender Perspective ..................................................... 74 D. Indicators of School Quality .......................................................................................... 80 E. Conclusions ................................................................................................................... 83 Chapter 7: Health and Poverty ................................................................................................... 84 A. Introduction ................................................................................................................... 84 B. Overview of the Health System ..................................................................................... 84 C. Health, Nutrition, and Population Status ....................................................................... 86 D. Access to and Utilization of Health Care....................................................................... 90 E. Health Sector Management, Financing, and Expenditure ............................................. 92 F. Other Determinants of Health Outcomes: Distance to Health Care, Sources of Water and Sanitation and Nutrition ............................................................................... 95 G. Conclusions ................................................................................................................... 96 H. Adult Equivalent Consumption ­ General Formula .................................................... 108 I. Adjustment of Scales Based on the Parameters of Modal Household Structure ......... 108 References ................................................................................................................................... 126 iii Acknowledgements This Poverty Assessment (PA) Report was prepared by a World Bank team as part of the Tajikistan Programmatic Poverty Assessment (PPA). Preparation of the Report was led by Oleksiy Ivaschenko (Task Team Leader, Economist, ECSHD). World Bank team members included: Anne Bakilana (Economist, ECSHD), Saodat Bazarova (Operations Officer, ECSHD), Alexander Danzer (Consultant, ECSHD), Gabriel Francis (Program Assistant, ECSHD), Ghaffar Mughal (Consultant, ECSHD), Ufuk Guven (Social Protection Specialist, ECSHD), Rasmus Heltberg (Senior Technical Specialist, SDV), Vladimir Kolchin (Consultant, ECSHD), Maria Laura Sanchez Puerta (Economist, HDNSP), Firuz Saidov (Consultant, ECSHD), Thomas Sohnesen (Consultant, ECSHD), Sowmya Srinivasan (Consultant, HDNSP), Juan Carlos Parra (Consultant, HDNDE), Diane Steele (Household Survey Coordinator, DECRG), and Sasun Tsirunyan (Consultant, ECSHD). Within the World Bank, the Tajikistan Living Standards Survey (TLSS) 2007 implementation was supported by Helen Shahriari (Senior Social Scientist, ECSSD), Richard Pollard (Senior Water and Sanitation Specialist, ECSSD), Jean Fares (Senior Economist, HDNSP), and Kinnon Scott (Senior Economist, DECRG). The Report also benefited greatly from the work of our editor, Diane Stamm. This Report would not be possible without the generous financial and technical support of the United Nations Children's Fund (UNICEF). We especially would like to acknowledge the support we received from Yukie Mokuo (UNICEF Representative, Tajikistan, until 2008), Hongwei Gao (UNICEF Representative, Tajikistan, since 2008), Ruth Leano (UNICEF Deputy Representative, Tajikistan), Farhod Khamidov (M&E Officer, UNICEF), Musharraf Solieva (Officer, UNICEF Tajikistan), and Yeva Gulnazaryan (Consultant, UNICEF). We would also like to acknowledge the generous financial support received from the Department of International Development (U.K.) (DFID) for TLSS 2007 implementation and dissemination of its results. In particular, we would like to acknowledge Dylan Winder (Head, DFID Tajikistan), Emily Poskett (Statistics Advisor, DFID), and Shahlo Rahimova (Project Officer, DFID). We have also received co-funding for TLSS 2007 implementation, and excellent technical input, from the Food and Agriculture Organization (FAO) partly via the EC-FAO Food Security Information for Action Program, and our special thanks go to Alberto Zezza (Economist, FAO). Zerkalo, the Center for Sociological Research, in Tajikistan, provided excellent input into the development of the data entry program and helped with the data entry, and we would like to acknowledge Qahramon Baqozoda (Director) and Ruslan Nabiev (Programmer). Aysholpan Dauletbaeva and Alisher Kholmatov provided excellent training for the fieldworkers. We extend many thanks to the State Statistical Committee of Tajikistan (GosKomStat) for its great work in preparation of the TLSS 2007, and in implementing the fieldwork. We are grateful to Mirgand Shabozov (Chairman, GosKomStat), Bakhtiya Mukhammadieva (First Deputy, GosKomStat), and Barot Turaev (Head of the Household Statistics Division, GosKomStat). During survey preparation we also benefited considerably from the comments and suggestions of representatives of various international organizations, including Don Van Atta (Policy Specialist, EU-TACIS Program), Robert Brudzynski (Team Leader, EC Budgetary Support Program), Mahmoud Naderi (Chief of Mission, IOM Tajikistan), Asif Niazi (Regional Assessment Officer, WFP), and Dilbar Turakhanova (ILO-IPEC Country Project Coordinator). The Report has benefited greatly from the peer reviewer comments (on both the Concept Note and the Draft Report) provided by Peter Lanjouw (Research Manager, DECRG), Branko Milanovic (Lead Economist, DECRG), Michael Mills (Lead Economist, AFTP1), and Quentin Wodon (Adviser, HDNDE). Very useful comments were also received from Sudharshan Canagarajah (Lead Economist, ECSPE), Annette Dixon (former Country Director, ECCU8), Elena Glinskaya (Senior Economist, ECSHD), Jariya Hoffman (Senior Economist, ECSPE), and Mehrnaz Teymourian (Country Program Coordinator, ECCU8). The task was undertaken under the guidance of Tamar Manuelyan Atinc (Sector Director, ECSHD), Arup Banerji (former Sector Manager, ECSHD), Gordon Betcherman (former Sector Manager, ECSHD Economics), Jesko Hentschel (Sector Manager, ECSHD Economics), Annette Dixon (former Country Director, ECCU8) and Motoo Konishi (Country Director, ECCU8). iv Executive Summary Despite Tajikistan's sustained economic growth of the past few years and the country's notable achievements, poverty and low standards of living remain a pressing problem for the majority of Tajik people. Poverty reduction is therefore one of the priority goals of the Government of the Republic of Tajikistan, of its national and international partners, and of Tajik society as a whole. This report presents a comprehensive analysis of poverty in Tajikistan using the most recent Tajikistan Living Standards Survey (TLSS) data, which is for 2007.1 The report provides an in-depth analysis of poverty dynamics (since 2003); a poverty profile; the linkages between labor markets, migration, and poverty; the importance of social transfers for poverty alleviation; and the key issues in the health and education sectors. The report also simulates possible poverty impacts from the reduction in migration and remittances related to the global financial crisis. The report shows that migration and remittances are a significant factor in the observed improvements in living standards between 2003 and 2007. However, the expected reduction in migration and remittances due to the financial crisis is likely to result in increased poverty levels, vulnerability, and higher income inequality, at least in the short term. The report has the following key findings: · Tajikistan achieved substantial welfare improvements between 2003 and 2007. Using the Purchasing Power Parity (PPP) US$2.15 per day poverty line, we find that poverty headcount declined from 64 percent in 2003 to 41 percent in 2007. Subjective measures of welfare and the data on possession of durable goods also confirm that, between 2003 and 2007, living standards in Tajikistan improved noticeably. · The welfare improvements between 2003 and 2007 were related to rising migration and remittances. The volume of remittances increased from about US$0.3 billion in 2004 to US$1.4 billion, or 40 percent of the country's gross domestic product (GDP), by the end of 2007. The analysis indicates that a substantial part of the overall observed reduction in the poverty headcount can be attributable to the direct or indirect impact of remittances. · Despite improvements in welfare since 2003, Tajikistan remains a country with widespread poverty. As mentioned, at the end of 2007, about 41 percent of the population was living under the poverty line of the PPP US$2.15 per day. Using the absolute poverty line derived from the TLSS 2007, based on the Cost of Basic Needs approach, we find that 53 percent of the population was poor and 17 percent was extreme poor (that is, living below the food poverty line). Seventy-five percent of the poor live in rural areas (as do 71 percent of the extreme poor). · The issue of the working poor continues to be one of the dominant features of poverty in Tajikistan. Half of the employed in the domestic labor market are poor. Almost 80 percent of the working poor live in rural areas. Low labor incomes and high prevalence of temporary work arrangements, informality (no labor contract), and unpaid work are the main reasons there are so many working poor. 1 The Tajikistan Living Standards Survey (TLSS) 2007 is described in Annex 1. 1 · A high household dependency ratio related to high fertility and the low labor market participation of women are substantial barriers to poverty reduction. The average fertility rate of three children per woman results in the very high dependency rate within a household, especially since women's participation in the labor market is low for various reasons. Among 2 million people who are out of the labor force, about 1 million are housewives. The analysis indicates that households with three or more children account for 53 percent of the total population, and 62 percent of all poor. The poverty risk in this group of households is above 60 percent compared with a 43 percent poverty risk for households with one child and 33 percent for households with no children. · Migration, mostly in the form of temporary work abroad, has become one of the key strategies for households to cope with poverty. The analysis indicates that a quarter of households have at least one migrant abroad. In households that have migrants, remittances account for as much as 35 percent of household consumption--and even more for the households in the lower deciles of the consumption distribution. The Tajikistan migration model is one of predominantly seasonal low-skill migration, with 96 percent of the migrants heading to Russia, and of those, 55 percent worked in the construction sector, and another 30 percent in other low-skill jobs. · The financial crisis is likely to increase poverty and inequality. The negative welfare impact of the financial crisis in Tajikistan is expected to be felt mostly through the reduction in migration and remittances. The conducted simulations indicate that in the short run the poverty headcount is likely to increase from 53 percent to at least about 58 percent. Moreover, poverty depth, poverty vulnerability, and consumption inequality are also expected to rise. · The social protection (SP) system in its current form is not positioned to be an effective instrument to reduce poverty. Against a background of high poverty and rising vulnerability in times of the financial crisis, Tajikistan runs a weak SP system dominated by old-age and disability pensions with little social assistance. Social assistance spending is very low (0.5 percent of GDP)-- the lowest in the Europe and Central Asia (ECA) Region--and programs are weak. As a result, the impact of the SP system on poverty and inequality is only marginal (reducing poverty headcount by 1.5 percentage points), but could be improved by a new, unified social assistance program offering poverty-targeted cash transfers. · The poor continue to be disadvantaged in terms of access to good-quality education services and health care. In the education sector, the key issues include generally poor and highly unequal quality of school facilities, across geographic areas, and the pervasive use of tuition fees in general education. In the health sector, the key issues include inadequate medical infrastructure and high out- of-pocket payments. However, in both sectors, some good reforms, especially those affecting financing and the incentives structure, are taking place. The key issues are discussed in greater detail below. A. Poverty Dynamics Tajikistan registered impressive rates of economic growth between 2003 and 2007. Real per capita GDP during this period is estimated to have increased by a cumulative 26.5 percent, or an average of 6.6 percent per year. As a reflection of this growth, average monthly per capita income (from National Accounts) increased in real terms from 119 somoni in 2003 to 150 somoni in 2007.2 2 US$1 = 3.4 somoni, thus 150 somoni is equal to US$44. 2 The analysis of the TLSS 2003 and 2007 data based on the comparable consumption aggregate indicates that during this period real per capita consumption increased significantly. The average rate of annual growth in real per capita consumption at the national level has been 7.5 percent.3 However, urban areas grew on average at a faster rate than rural areas, with 2003­07 growth rates of 9.9 percent and 6.6 percent, respectively. As a result, the gap in average consumption between urban and rural areas increased between 2003 and 2007. Economic growth during 2003-07 was pro-poor. The analysis indicates that lower deciles of the per capita consumption distribution have registered higher rates of growth. Indeed, the per capita consumption of the poorest 20 percent (1st quintile) of the population grew at an average annual rate of 13.6 percent, compared to the 8.3 percent for the 3rd (medium) quintile of the distribution, and 5.7 percent for the richest quintile (Table 1). Table 1: Changes in Real Per Capita Consumption across Quintiles, 2003­07 % change, % change, 2003 2007 cumulative annual Lowest quintile 47.8 74.0 54.6 13.6 2 74.6 106.4 42.5 10.6 3 100.1 133.2 33.1 8.3 4 134.9 169.8 25.8 6.5 Highest quintile 245.7 301.5 22.7 5.7 Total 120.6 156.9 30.1 7.5 Source: World Bank estimates based on the TLSS 2003 and 2007. Using the absolute poverty line derived from the TLSS 2007, we find that the poverty headcount declined from 72 percent in 2003 to 53 percent in 2007.4 This reduction in poverty implies a poverty- growth (in per capita consumption) elasticity of -1. In other words, a 1 percent growth in per capita consumption has resulted in a 1 percent decline in the poverty headcount.5 Using this poverty line, we find that during this period the poverty headcount declined from 74 percent to 55 percent in rural areas and from 69 percent to 49 percent in urban areas (Figure 1). About 1.05 million people escaped absolute poverty between 2003 and 2007, despite the increase in the total population by an estimated 0.4 million between those years (to an estimated 7.1 million in 2007). The incidence of extreme poverty declined from 42 percent in 2003 to 17 percent in 2007. The decline in extreme poverty6 has been more rapid in rural areas. As a result, as of end-2007, rural areas had a somewhat lower incidence of extreme poverty relative to urban areas (16.4 percent compared to 18.9 percent), while the incidence of total poverty was higher in rural areas (55 percent compared to 49.4 percent). The story of the improvement in household welfare during 2003­07 is also supported by non- consumption welfare indicators. Fifty-five percent of the population lives in households where household heads believe their financial situation improved over the last three years. Importantly, among the households that are quantified as consumption poor, this share is also quite high, at 48 percent. Also, between 2003 and 3 This number is based on comparing the consumption from the household surveys (with the data adjustments done as highlighted above), and hence it is not directly comparable to the growth rate in per capita consumption obtained from the National Accounts, although both numbers are very close. 4 This absolute poverty line was derived from the TLSS 2007 based on the Cost of Basic Needs (CBN) approach. The previous poverty assessments, which were based on the TLSS 1999 and 2003, used the PPP US$2.15 poverty line. The absolute poverty line derived in 2007 is equal to 139 somoni per month (while the value of the PPP US$2.15 poverty line is equal to 120 somoni in 2007 prices). 5 The PPP US$2.15 per day poverty headcount declined from 64 percent in 2003 to 41 percent in 2007. We report the PPP US$2.15 poverty estimates for comparability purposes, since the same poverty line was used in 2003. 6 The extreme poverty line is effectively the food poverty line, since it reflects the cost of the reference group typical food basket needed to get 2,250 calories per person per day. The value of the extreme poverty line is equal to 89 somoni per month in end-2007 prices. 3 2007, the possession of such "modern" durable goods as color TVs, cars, DVD players, and so forth has increased in both urban and rural areas, with the higher rate of increase in rural areas and among the poor. Figure 1: Changes in the Poverty Headcount (TLSS 2007 absolute poverty line), 2003­07 80 03' 70 73.8 72.4 07' Poverty headcount, % 60 68.8 50 55.0 53.5 49.4 40 30 20 10 0 urban rural total Source: World Bank estimates based on the TLSS 2003 and 2007. The increase in migration and in related remittances is an important factor behind the poverty reduction between 2003 and 2007. Simulations carried out for this report indicate that had remittances in 2007 remained at their 2003 level, the incidence of both total and extreme poverty would be much higher compared to the actually observed 2007 poverty levels. While migration and remittances are a significant part of the story behind the 2003­07 economic growth and poverty reduction in Tajikistan, they are also a source of vulnerability in times of the financial crisis. This is due to the nature of Tajik migration, whereby about 96 percent of migrants go to Russia (and out of those more than 51 percent chose Moscow as their destination), and 55 percent work in construction. Since the construction boom in Russia has effectively come to a halt, a sharp reduction of demand for imported labor is unavoidable. Simulations on the poverty impact of the reduction in remittance flows to Tajikistan confirm the country's substantial external dependence.7 We find that a 30 percent decline in remittances is expected to increase the poverty headcount from 53.1 percent to 57.9 percent.8 Rural areas are expected to be more strongly affected than urban areas due to a higher concentration of migrants there. The decline in remittances/employment abroad would also lead to an increase in poverty depth (that is, the already poor will become poorer), vulnerability to poverty (that is, non-poor moving closer to the poverty line), and income/consumption inequality. B. Poverty Profile A significant reduction in poverty has taken place in Tajikistan since 2003, but poverty still remains widespread and deep. There is also a noticeable clustering of the population around the poverty line. The important policy implication of clustering is that even a relatively modest change in the purchasing power of the population in Tajikistan will induce significant changes in the prevalence of poverty. Indeed, a 7 All effects are pure direct remittances effects, without taking into consideration the potential economic recession in Tajikistan (that is, worsening conditions in the local labor market) and without accounting for potential second-order effects from remittances (multiplier effect). 8 According to recent data on the flow of remittances reported by the National Bank of Tajikistan, remittances have already declined by 20 percent compared to the respective period last year, and are expected to decline further. 4 10 percent decline in purchasing power would increase the poverty headcount from 53.5 percent to 62.6 percent, or by almost 10 percentage points (Table 2). In the current environment of financial crisis such a scenario is increasingly realistic. Table 2: Sensitivity of Headcount Poverty Rate with Respect to the Choice of Poverty Line, 2007 Poverty Change from actual Poverty Change from actual Change in the poverty line Incidence (P0) (%) Incidence (P0) (%) Poverty Line = Total (139 Somoni) Poverty Line = Extreme (89 Somoni) Actual 53.5 0.0 17.1 0.0 +5% 58.6 9.5 20.2 18.0 +10% 62.6 16.9 23.3 36.1 +20% 69.4 29.8 29.7 74.0 -5% 49.0 -8.4 14.4 -15.7 -10% 44.5 -16.8 12.0 -29.5 -20% 33.8 -36.9 7.4 -56.9 Source: World Bank estimates based on the TLSS 2007. Poverty continues to be concentrated in rural areas. In Tajikistan, 75.7 percent of all poor (and 70.9 percent of extreme poor) live in rural areas, reflecting demographics (73.7 percent of the overall population live in rural areas) and higher poverty incidence in rural areas. The main correlates of poverty include: (a) high dependency ratios related to high fertility, (b) employment in low-pay agriculture, (c) low level of education, and (d) unfavorable geographic conditions, such as a high altitude. To cope with poverty, households resort to sending temporal migrants abroad. These factors are elaborated upon below. High household dependency rates in Tajikistan continues to be one of the key correlates of the high poverty risk. Households with seven or more household members account for 58.1 percent of the total population and 66.5 percent of the total poor. Higher poverty rates among larger households are related to higher dependency ratios. In fact, Tajikistan has the highest total dependency rate--defined as the number of people under age 15 plus the number of people age 65 and older per 100 persons 15 to 64 years old--in the ECA region (Figure 2). High total dependency in Tajikistan is driven almost exclusively by high child dependency, as elderly dependency in Tajikistan is lowest in the ECA region. Figure 2: Total Dependency Rates in Tajikistan Compared to other ECA Countries, 2005 80 70 60 50 40 30 20 10 0 Turkmenistan Romania Ukraine Uzbekistan Tajikistan Turkey Czech Russian Poland oldova Estonia Kazakhstan Croatia Georgia Montenegro Slovak Hungary Macedonia, Slovenia Bosnia Bulgaria Latvia Lithuania Serbia Armenia Belarus Kyrgyz Azerbaijan Albania M Source: United Nations World Population Prospects. 5 The story of poverty in Tajikistan is also one of the working poor. Employment in agriculture is not associated with a reduced risk of poverty. Those households where the household head is wage-employed in agriculture face a higher risk of poverty than the national average--58.3 percent and 16.3 percent for total and extreme poverty, respectively (Figure 3). Self-employment in agriculture is associated with a high risk of poverty as well. The high risk of poverty for those employed in agriculture is especially worrisome considering that agriculture accounts for 45 percent of the total domestic employment in Tajikistan. Figure 3: Poverty Headcount by the Status of Employment Total povety, % 70 60.4 Extreme poverty, % 58.3 60 52.7 53.5 47.8 50 44.8 40 30 21.0 18.9 17.1 20 13.2 16.3 13.7 10 0 Not employed Wage employed Wage-employed Self -employed Self -employed Total (non-agr.) (agr.) (non-agr.) (agr.) Source: World Bank estimates based on the TLSS 2007. Households with more educated heads are much less likely to be poor. Differences in educational attainment of heads of households are reflected in considerably different poverty rates. The risk of total poverty increases from 37 percent for households with a head with a higher (university) education to above 60 percent for households with a head with a basic (or below) level of education. Altitude is highly correlated with poverty levels. In Tajikistan, many settlements are located at very high altitudes, where weather conditions are extremely harsh and the main source of gainful employment comes from temporal migration. The correlation between altitude and poverty can be clearly seen in the mountainous region of Gorno-Badakhshan Autonomous Oblast (GBAO), where as a result of mostly those pronounced differences in altitude, the poverty headcount across areas varies from 20 percent to 80 percent depending on the altitude. Many households send their members to work abroad as a poverty-coping strategy. Migration (employment abroad) does reduce the risk of poverty. The effect is especially evident when one looks at the share of migrants in the total household size, not simply the number of migrants. We find that households in which the share of migrants exceeds 20 percent have total and extreme poverty headcounts of 42.6 percent and 13 percent, respectively, compared to 54.4 percent and 17.4 percent, respectively, for households in which the share of migrants is under 20 percent. As a reflection of widespread poverty, food insecurity continues to be a problem for many households. At the time of the household survey (October­November 2007), 24 percent of individuals were living in households with inadequate (self-assessed) food consumption. In the bottom quintile of consumption distribution this share rises to 44 percent. Seasonality in food security affects all welfare groups, and most severe shortages of food are experienced during late winter-early spring (Figure 4). 6 Figure 4: Seasonality in Food Security, by Welfare Groups (percent of food-secure) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% a- 7 u-6 e-6 o- 6 a- 7 Mr 7 Ar 7 u- 7 u-7 c 6 e- 6 Fb 7 u 7 My 0 A g0 S p0 N v0 a-0 p-0 A g0 Ot-0 D c0 J n0 e -0 J n0 J l-0 Extr e m e Poor Non poor Total Source: World Bank estimates based on the TLSS 2007. C. Domestic Labor Market Tajikistan has a population of 7 million, of which the working-age (15­64 years old) population comprises 4.2 million people. This is further divided almost equally between those who are a part of the labor force (2.2 million) and those who are out of the labor force (2 million). The number of employed is 1.95 million, which constitutes 47 percent of the working-age population. The unemployment rate is 9.5 percent. Key issues in the domestic labor market include: (a) low participation, especially among women; (b) dominance of agriculture as the main sector of employment; (c) low labor earnings, especially in the agricultural sector; and (d) high prevalence of informal job arrangements. Labor market participation in Tajikistan is much lower than the ECA average. More than 50 percent of the economically inactive population is housewives. The labor force participation rate is extremely low at 52 percent. This is mostly attributed to the low female labor force participation rate at 36 percent, which is 30 percentage points lower than the male participation rate in Tajikistan. There are close to 1.1 million housewives in Tajikistan. These women are primarily concentrated in the 25­54 age group and rural areas. Agriculture and related activities dominate the employment sector in Tajikistan. This sector accounts for 45 percent of the employment sector in the country. The next-most-important sectors of employment are services and commerce, together accounting for 30 percent of total employment. Employment rates increase with education for workers of all age groups. Labor incomes are the lowest in the agriculture sector, even after accounting for income from agriculture and value of in-kind consumption. At 80 somoni a month, median earnings in the agriculture sector are the lowest among all sectors of employment, and are much lower than the total poverty line of 139 somoni. The services sector has the next-lowest median monthly income at 159 somoni. The construction sector earns the highest median monthly income at 366 somoni (Figure 5). 7 Figure 5: Median Monthly Labor Incomes by Sector of Employment Source: World Bank estimates based on the TLSS 2007. More than one-third of workers are in the informal sector.9 The informal part of the economy is dominated by the agriculture and services sectors. Forty-seven percent of workers employed informally are in the agriculture sector, and 25 percent are in the services sector. The income premium for working in the formal sector is 10 percent once other determinants of income are taken into account. D. Migration After the collapse of the Former Soviet Union, both internal and external migration in Tajikistan increased sharply. While internal migration was mainly caused by a severe civil war between 1992 and 1997, external migration was driven by ethnic motivations in the first years of independence, but became dominated by labor motives soon after. In terms of economic importance, external migration plays a dominant and increasing role, and is thus the key focus of the analysis presented here. The TLSS 2007 data contain very rich information on migrants (Box 1). Box 1: What We Know and Do Not Know about Migrants from the TLSS 2007 The TLSS 2007 has comprehensive migration modules covering both internal and external migration. With respect to external migrants, the information covers both return migrants and "current" migrants--those who are reported to be working abroad at the time of the survey. When it comes to both groups of external migrants, the information about them includes: (a) migration destination (country and city), (b) duration of the last migration spell, (c) exact months of exiting and reentering Tajikistan, (d) sector of employment, (e) monthly wage, and (f) legal status. The limitations of the migration modules in the TLSS 2007 are that they collect detailed information only on the last migration spell and provide only limited information on past migration histories. Thus, we certainly underestimate past (and total) migration flows. The main issues arising from the analysis of migration in Tajikistan are: (a) a very high dependence of the country on external migration and remittances; (b) the low-skill nature of migration, which is predominantly directed towards the construction sector in Russia; and (c) high reliance of households, and especially the poor ones, on remittances as a source of income. 9 In this paper, informality is defined as the lack of a signed contract with the employer in the primary job. 8 According to official data, the inflow of remittances to Tajikistan in 2007 amounted to US$1.4 billion, or about 40 percent of the country's GDP, the highest in the world. Moreover, the volume of remittances has increased sharply since 2003, when it was around US$0.3 billion. Russia is the main sending country of remittances, accounting for 80 to 90 percent of Tajikistan's officially recorded remittances. The TLSS 2007 estimated the number of migrants to be 350,000 in 2007 (January­October). Of those, about 100,000 had already returned in 2007. Five percent of the population of Tajikistan was abroad in 2007, which equals 19 percent of the working population. However, the TLSS 2007 data is likely to underestimate total migration flows (see Box 1). About 96 percent of migrants chose Russia, and out of those more than 51 percent chose Moscow as their destination. About 55 percent of migrants work in the construction sector. Another 30 percent work in other low-skill jobs (trade, services, and so forth). Ninety-nine percent of migrants are male. The median migration spell of return migrants is about eight months. Around 75 percent of migrants hold a secondary education degree. Migrants on average earn about US$300 per month. Remittances have a significant bearing on household consumption. On average, remittances account for 10 percent of what urban households can buy and 15 percent of what rural households can buy, on annual basis. When conditioning on remittances receipt, the substantial depth of external dependence becomes even clearer. The share of annual consumption for those households that receive remittances exceeds 35 percent in all welfare quintiles. The poorest rural and urban households which receive remittances finance, on average, almost 80 percent and 50 percent, respectively, of their annual consumption through remittances. E. Social Protection Against a background of high poverty and low employment, Tajikistan runs a rudimentary social protection (SP) system dominated by old-age and disability pensions with virtually no social assistance. By far, the largest program in terms of coverage is the old-age pension, which is received by one-third of households. Total social assistance spending is very low--at 0.5 percent of GDP it is the lowest in the ECA Region--and programs are small in size and benefit coverage. Less than 1 percent of households receive any of the smaller social assistance benefits, such as the gas and electricity compensation. Given its small size it is not surprising that the SP system has very small impact on poverty and inequality. Our estimations suggest that transfers reduce poverty headcount from 54.8 percent to 53.1 percent, a reduction of 3 percent, or 1.7 percentage points. The estimated reduction in the poverty gap is 5.1 percent. The estimated impact of social transfers on inequality is also very low, suggesting a reduction in the country's already low Gini Coefficient of inequality of 1.5 percent.10 To improve the SP system and its impact on poverty, one option would be to introduce a targeted social assistance to achieve a higher coverage of the poor and vulnerable. A useful starting point would be to consider a single, unified social assistance program of some scale with competent staff trained in targeting, delivery, and complaints handling. Expanding social assistance does entail short-term fiscal costs, but this would help fight poverty and offer citizens a protection mechanism against a range of shocks and crises. The main targeting method to consider for the proposed benefit is proxy means testing based on a formula linking observable factors to expected well-being and community-based targeting. 10 These estimations assume that all transfers are consumed and that transfers have no offsetting value on labor supply and other income sources. 9 F. Education Tajikistan has made progress in education, but the country still faces major challenges in improving education outcomes. Tajikistan managed to sustain high enrollment ratios in primary and secondary education. Both public and private funding for education also increased significantly to support this trend. However, many challenges remain. The main issues in the education sector include: (a) very poor school infrastructure, including access to electricity and heating in winter months (resulting in temporary school closures); (b) perceived lack of qualified teachers especially in rural areas, while student-teacher ratios are comparatively low; (c) hidden costs and/or tuition fees, which represent a substantial burden for poor households; and (d) public financing arrangements relying significantly on the budgets of local authorities, which introduces inequalities in the quality of education services across regions. The quality of school facilities is often deficient, with large inequalities across regions. Rural and poor schools are particularly affected by poor infrastructure and they also have seen the least improvement in facilities over the last five years. Over 40 percent of rural areas report their school facilities to be of unsatisfactory quality in all aspects. At the same time, there is evidence that improvements might not have taken place in areas that have the greatest need. The reported lack of qualified teachers and low student-teacher ratios point to institutional challenges. The lack of teachers seems to be related more to location specifics and low pay than to the overall wealth of the location. At the same time, the student-teacher ratios are comparatively low, indicating that the major challenge is to adjust the institutional setup to use and allocate teachers optimally. Adjustment of curriculums, promotion of multi-subject and multi-grade teachers, potentially combined with payment incentives for rural schools, should be looked into further. The use of tuition fees or hidden charges in general education represents a significant problem for poor households. The analysis presented in this report indicates that a substantial share of children in public general education pays tuition fees or charges, despite the Government's policy of offering free education. The practice of charging fees is most common in Dushanbe. There is a clear link between the prevalence of school fees and the percentage of children not attending school for financial reasons. Ensuring free education for all is important in achieving universal general education. Most schools are locally financed and are therefore dependent on funding and decisions of local governments, which are accountable to the central government to a limited degree only. This system results in unequal funding of schools, hinders oversight, and provides no incentives for efficiency at the school level. Current reforms are addressing some of these issues. Per capita funding, budget reforms, and more autonomy in schools are likely to contribute to better oversight and efficiency. The impact of those reforms needs to be assessed as more data and evidence become available. G. Health The health status of Tajik population continues to be among the worst in the ECA Region. The poor continue to be much more disadvantaged in terms of access to health care, and they have worse health outcomes compared to non-poor. In Tajikistan, the gap in health outcomes is evident through much higher rates of child morbidity and mortality in poor households. Despite having poorer health, the poor have comparatively less access to health care services. Tajikistan's health system is not pro-poor, depends on out- of-pocket payments, and is not equitable. Nor does it offer financial protection from expensive catastrophic illness. The TLSS 2007 reveals that almost 46 percent of the poorest households found it impossible or very difficult to pay for health care compared to 28 percent of the richest households. 10 Management, financing, and expenditure weaknesses affect the health care system and they have a greater impact on the poor. Low public financing, excessive and outdated hospital infrastructure, underfunded primary health care, and inefficiency in resource allocation all contribute to poor health outcomes. Without reforms to improve the efficiency with which meager resources are used, health outcomes will continue to be poor. The Government began reforming the sector as early as the 1990s and the Ministry of Health continues to work closely with development partners, for example, in implementing a Health Financing Strategy, which aims to tackle financing, allocation, and ways that health care providers are paid. In the last few years, the Government has started to consistently pursue some of the recommended reforms needed in the health sector. The Government has introduced changes that aim to decrease the proportion of the population for who health care is unaffordable, to decrease the share of health costs in household expenditure, and to improve equity in the distribution of the health budget by introducing per capita financing for primary health care and case-based payments for hospitals. However, despite the Government's increased allocations to the health sector, the additional resources are not sufficient to meet the needs of the sector. The latest TLSS 2007 shows that the bulk of health care expenditure is still borne by households. The cost of health care continues to be a deterrent to accessing care, especially by the poor. Nevertheless, due to improvements in household welfare since 2003, more people do seek health services. To achieve further progress, the population needs to be better informed of the costs of care and the choices they have in terms of where to access care. Such practices are being piloted. Additional surveys will be needed to be able to measure the full impact of reforms that have been gradually introduced in the last few years. H. Summary Tajikistan achieved significant progress in the reduction of poverty between 2003 and 2007. However, poverty remains widespread and linked to specific characteristics. Those characteristics include high fertility, concentration of the poor in rural areas, poor domestic labor market conditions (which result in a high dependence on remittances), and a weak social protection system dominated by old-age and disability pensions with virtually no social assistance. The poor continue to be disadvantaged in terms of access to good-quality education services and health care. Moreover, the ongoing financial crisis represents a real threat to poverty reduction efforts. There is a real risk that poverty and inequality would be on the rise again, since migration and remittances are identified as the main transmission channels of the global financial crisis to Tajikistan. The Government of Tajikistan needs to monitor the remittances situation and to think about the labor policy instruments that could absorb the potential increase in the supply of return migrants in the domestic labor market, such as public works or large infrastructure projects. 11 Chapter 1: Poverty Dynamics in Tajikistan11 Introduction The newly collected household survey data from the 2007 Tajikistan Living Standards Survey (TLSS 2007) provide an opportunity to establish the evolution of poverty since the last survey (2003). The data also provide a good baseline for simulating the impact of the ongoing financial crisis on poverty. Hence, this chapter will discuss the following key aspects: (a) economic growth and consumption-based poverty dynamics between 2003 and 2007, and (b) the expected poverty impact of the financial crisis. While the analysis presented here establishes the link between economic growth and poverty, it does not discuss the sectoral patterns and drivers of economic growth, since this issue is addressed in detail in the Country Economic Memorandum (CEM). Economic Growth and Trends in Consumption between 2003 and 2007 Tajikistan registered impressive rates of economic growth between 2003 and 2007. Real gross domestic product (GDP) during this period is estimated to have increased by a cumulative 36.1 percent, or an average of 9 percent per year. However, considering the estimated increase in population, this growth translates to an annual rate of economic growth per capita of 6.6 percent, still a very solid number. As a reflection of this growth, average monthly per capita income (from National Accounts) increased in real terms from 119 somoni in 2003 to 150 somoni in 2007.12 And since consumption in National Accounts (NA) represents about 75 percent of the GDP, the respective monthly per capita consumption numbers are 89 somoni and 113 somoni, respectively (Table 1.1). Table 1.1: Tajikistan Macro Trends, 2003­07 2003-07 2003-07 avg. cumulative change per 2003 2007 change, % annum, % Population, million 6.6 7.1 7.6 1.9 Annual GDP (in constant 2007 prices, million Somoni) 9,388 12,780 36.1 9.0 Annual GDP per capita (in constant 2007 prices, Somoni) 1,422 1,800 26.5 6.6 Monthly GDP per capita (in cons. 2007 prices, Somoni) 119 150 26.5 6.6 Monthly HH consumption per capita, NA (in cons. 2007 prices) 89 113 27.1 6.8 Source: National Accounts. The TLSS 2007, which was fielded during October­November 2007, presented an opportunity to establish the consumption and poverty trends since 2003. However, several methodological considerations had to be taken into account to make the comparisons possible. To make the consumption data comparable over time, the following adjustments have been made: (a) a 106.9 percent increase in prices 11 This chapter was written by Oleksiy Ivaschenko, Economist, ECSHD. Valuable support was provided by Sasun Tsirunyan, Consultant, ECSHD. 12 All numbers are expressed in constant 2007 prices. 12 between June 2003 and 2007;13 (b) a 16.4 percent increase due to the various times of the year the survey was conducted (since the 2003 survey was conducted mostly in June); this adjustment factor was derived empirically based on the monthly data from the 2005 panel of the Household Budget Survey households by comparing the value of November consumption to that of June consumption in real terms; and (c) a 6.0 percent increase due to using a more extensive list of food consumption items in the 2007 survey (this adjustment factor was derived by comparing the "full" consumption aggregate to the "restricted" [same items as in 2003] consumption aggregate). By combining these three adjustment factors we get an overall adjustment factor of a 155.3 percent increase. The growth rates in real per capita consumption have been unequal across regions. The average rate of growth in real per capita consumption at the national level has been 7.5 percent.14 Note that this number is based on comparing the consumption from the household surveys (with the data adjustments done as highlighted above), and hence it is not directly comparable to the rate obtained from the National Accounts, although both numbers are fairly close. Urban areas grew on average at a faster rate than rural areas, with the 2003­07 growth rates of 9.9 percent and 6.6 percent, respectively. As a result, the gap in average consumption between urban and rural areas increased between 2003 and 2007. Looking across oblasts (regions), Gorno-Badakhshan Autonomous Oblast (GBAO) registered the highest rate of growth, although from the lowest level back in 2003. This oblast, however, accounts for only 3 percent of the total population. The lowest rate of growth happened in the Region of Republican Subordination (RRS) oblast, where per capita consumption grew at 3.4 percent per year (Table 1.2). Table 1.2: Changes in Real Per Capita Consumption across Regions, 2003­07 % change, % change, 2003 2007 cumulative annual Total 120.6 156.9 30.1 7.5 Urban 129.2 180.2 39.5 9.9 Rural 117.4 148.6 26.6 6.6 Oblast (region) Dushanbe 151.1 183.0 21.1 5.3 Sogd 118.6 146.7 23.6 5.9 Khatlon 100.8 154.7 53.4 13.4 RRP 143.0 162.1 13.4 3.4 GBAO 86.2 164.7 91.2 22.8 Source: World Bank estimates based on the 2003 and 2007 TLSS. The good news is that growth has been pro-poor. The analysis indicates that lower quintiles of the per capita consumption distribution have registered higher rates of growth. Indeed, the per capita consumption of the poorest 20 percent (1st quintile) grew at an average annual rate of 13.6 percent, compared to the 8.3 percent for the 3rd (medium) quintile of the distribution, and 5.7 percent for the richest quintile (Table 1.3). 13 The GDP deflator over the same period suggests an increase in prices of 105.6 percent, which is very close. To calculate the overall Consumer Price Index (CPI), for its food component we use the food price index derived directly from the household survey, which indicates that food prices between 2003 and 2007 increased by 141 percent (while official food CPI suggests an increase of 60 percent). 14 The growth incidence curves at the national and urban/rural levels are presented in Annex 7, Figure A7.1. 13 Table 1.3: Changes in Real Per Capita Consumption across Quintiles, 2003­07 % change, % change, 2003 2007 cumulative annual Lowest quintile 47.8 74.0 54.6 13.6 2 74.6 106.4 42.5 10.6 3 100.1 133.2 33.1 8.3 4 134.9 169.8 25.8 6.5 Highest quintile 245.7 301.5 22.7 5.7 Total 120.6 156.9 30.1 7.5 Source: World Bank estimates based on the 2003 and 2007 TLSS. Driven by robust growth in consumption, poverty declined significantly. The Purchasing Power Parity (PPP) US$2.15 per day poverty headcount declined from 64 percent in 2003 to 41 percent in 200715 (Table 1.4). Interestingly, despite the fact that the rate of increase in per capita consumption was lower for rural areas, the poverty headcount there declined in absolute terms (percentage points) somewhat more there, because of the more pronounced poverty in rural areas. This decline in the poverty headcount in terms of the absolute number of the poor translates into 1.35 million people escaping poverty (with 1 million of those in rural areas), despite the increase in the total population by an estimated 0.4 million between 2003 and 2007. Table 1.4: Changes in the Poverty Headcount (PPP US$2.15 poverty line), 2003­07 Poverty headcount, % N of the poor Population 2003 (actual) national 63.5 4,233,341 6,667,510 urban 59.1 1,066,293 1,804,290 rural 65.1 3,167,048 4,863,220 2007 (actual*) national 40.9 2,886,561 7,061,429 urban 40.3 747,778 1,856,865 rural 41.1 2,138,783 5,204,564 Absolute change (2007 vs. 2003) national -22.6 -1,346,780 393,919 urban -18.8 -318,515 52,575 rural -24.0 -1,028,265 341,344 Implied poverty-growth elasticity, % -1.34 Source: World Bank estimates based on the 2003 and 2007 TLSS. Using the absolute poverty line derived from the TLSS 2007, we find that the poverty headcount declined from 72 percent in 2003 to 54 percent in 2007.16 Using this poverty line, we find that during this period, the poverty headcount declined from 74 percent to 55 percent in rural areas, and from 69 percent to 49 percent in urban areas. Under this poverty line, about 1.05 million people escaped absolute poverty between 2003 and 2007 (Table 1.5). 15 We report here the PPP US$2.15 poverty estimates for comparability purposes, since the same poverty line was used in 2003. 16 The absolute poverty line was derived based on the Cost of Basic Needs approach, and is equal to 139 somoni per month (while the value of the PPP US$2.15 poverty line is equal to 120 somoni). The derivation of this poverty line is discussed in greater detail in Chapter 2 on the poverty profile. 14 Table 1.5: Changes in the Poverty Headcount (absolute poverty line), 2003­07 Poverty headcount, % N of the poor Population 2003 national 72.4 4,830,408 6,667,510 urban 68.8 1,241,352 1,804,290 rural 73.8 3,589,056 4,863,220 2007 national 53.5 3,779,802 7,061,429 urban 49.4 917,291 1,856,865 rural 55.0 2,862,510 5,204,564 Absolute change (2007 vs. 2003) national -18.9 -1,050,606 393,919 urban -19.4 -324,060 52,575 rural -18.8 -726,546 341,344 Implied poverty-growth elasticity, % -1.0 Source: World Bank estimates based on the 2003 and 2007 TLSS. The incidence of extreme poverty declined from 42 percent in 2003 to 17 percent in 2007. The decline in extreme poverty17 has been more rapid in rural areas. As a result, as of end-2007, rural areas had a somewhat lower incidence of extreme poverty relative to urban areas, while the incidence of total poverty was higher in rural areas (Table 1.6). Table 1.6: Changes in the Poverty Headcount (absolute poverty line), 2003­07 Headcount Rate(P0) Poverty Gap(P1) 2003 2007 change 2003 2007 change Poverty Line = Total (139 Somoni) Urban 68.8 49.4 -19.4 27.8 15.4 -12.3 Rural 73.8 55.0 -18.8 29.6 14.9 -14.7 Total 72.4 53.5 -18.9 29.1 15.0 -14.1 Poverty Line = Extreme (89 Somoni) Urban 39.4 18.9 -20.5 12.5 4.3 -8.2 Rural 42.3 16.4 -25.9 12.4 3.1 -9.3 Total 41.5 17.1 -24.4 12.4 3.4 -9.0 Source: World Bank estimates based on the 2003 and 2007 TLSS. The decomposition of the poverty changes into growth and inequality components indicates that most of the poverty reduction is related to growth. Indeed, of the total 18.9 percentage point decline in poverty, 17.4 points is attributable to the impact of economic growth (Table 1.7). Income inequality played only a very moderate counterbalancing role.18 It is widely believed that strong economic growth during that period was fueled by a sharp increase in remittances. Indeed, remittances have been identified as the main factor behind economic growth and the resulting reduction in poverty. 17 The extreme poverty line is effectively the food poverty line, since it reflects the cost of the reference group typical food basket needed to get 2,250 calories per person per day. The value of the extreme poverty line is equal to 89 somoni per month in end-2007 prices. 18 The rest of the impact is due to unexplained "residual" or "error" term in the regression model. 15 Table 1.7: Growth and Redistribution Decomposition of Poverty Changes Change in incidence of poverty actual Redistribu 2003 2007 Growth Interaction change tion Poverty Line = Total (139 Somoni) Total 72.4 53.5 -18.9 -17.4 2.0 -3.6 Urban 68.8 49.4 -19.4 -22.2 2.4 0.4 Rural 73.8 55.0 -18.8 -15.4 1.7 -5.1 Poverty Line = Extreme (89 Somoni) Total 41.5 17.1 -24.4 -16.8 -4.5 -3.1 Urban 39.4 18.9 -20.5 -19.6 2.7 -3.6 Rural 42.3 16.4 -25.9 -15.3 -7.2 -3.3 Source: World Bank estimates using LSMS 2007 data. Non-consumption indicators of the living standards also indicate that welfare of the population improved during 2003­07. Fifty-five percent of the population lives in households where household heads believe that the financial situation improved over the last three years. Importantly, among the households that are quantified as consumption poor, this share is also quite high--48 percent. Only 14 percent of the poor population reported that their financial situation deteriorated over the last three years (Figure 1.1) Figure 1.1: Changes in the Self-reported Financial Status over the Last Three Years 70% 63.3% 60% Improved 55.3% 48.3% 50% Same Deteriorated 40% 31.8% 36.2% 30% 26.6% 20% 11.3% 13.7% 8.6% 10% 0% Total Poor Non-poor Note: 2003/2007- actual level of poverty in 2003/07. 07* - Simulated level of poverty in 2007 had remittances stayed at the 2003 level. Source: World Bank estimates based on the 2003 and 2007 TLSS. The improvement in household welfare can also be seen from the increased possession of durable goods. The possession of "modern" durable goods19 between 2003 and 2007 increased in both urban and rural areas, with a higher rate of increase in rural areas. For instance, during this period, in urban areas the possession of color TVs increased from 44 percent to 77 percent and in rural areas from 15 percent to 53 percent (Table 1.8). 19 We call "modern" those durable goods the possession of which tends to increase with the household economic status, such as a color TV, computer, and car. We label as "traditional" those durable goods the possession of which tends to stay stable or decline with the household economic status, such as a refrigerator or a kerosene or wood stove. 16 Table 1.8: Changes in the Possession of Durable Goods (by urban/rural), 2003­07 Percentage Percentage point point Durable good 2003 2007 change, % 2003 2007 change, % urban urban rural rural A. "Modern" goods Color TV 44.4% 76.7% 32.3% 14.9% 53.1% 38.2% Video player 16.6% 24.8% 8.2% 6.6% 10.5% 3.9% Microwave owen 3.9% 11.1% 7.2% 1.7% 7.2% 5.5% Computer 1.1% 5.0% 3.9% 0.1% 0.6% 0.5% Satellite dish 1.4% 12.4% 11.0% 0.3% 4.0% 3.7% Car 11.0% 14.6% 3.6% 10.8% 16.5% 5.7% B. "Traditional" goods Refrigerator 53.5% 61.4% 7.9% 19.7% 20.2% 0.5% Wood stove 28.5% 28.2% -0.3% 82.8% 75.8% -7.0% Kerosine stove 2.3% 1.6% -0.7% 3.0% 2.2% -0.8% Radiator electric 24.4% 3.4% -21.0% 6.1% 1.4% -4.7% Black & white TV 46.7% 23.8% -22.9% 57.0% 44.6% -12.4% Source: World Bank estimates based on the 2003 and 2007 TLSS. We also find that poorer households accumulated "modern" durable goods at a higher rate (although from a much lower base), which is consistent with the consumption-based story of welfare improvements. Indeed, possession of color TVs by the poorest quintile increased from 10 percent to 46 percent, or by 370 percent, the highest rate of increase across quintiles (Table 1.9). Table 1.9: Changes in the Possession of Color TVs (by quintile), 2003­07 Percentage point change, Durable good Quintile 2003 2007 % A. "Modern" goods Color TV total 24.9% 60.7% 35.8% Q1 9.8% 46.1% 36.3% Q2 13.8% 55.5% 41.7% Q3 21.7% 57.0% 35.3% Q4 34.7% 65.6% 30.9% Q5 44.6% 72.7% 28.1% B. "Traditional" goods Black & white TV total 53.5% 37.9% -15.6% Q1 54.1% 48.0% -6.1% Q2 59.4% 42.4% -17.0% Q3 58.9% 41.5% -17.4% Q4 51.6% 34.2% -17.4% Q5 43.4% 28.5% -14.9% Source: World Bank estimates based on the 2003 and 2007 TLSS. The increase in migration and in related remittances is the most important factor behind poverty reduction during 2003­07. According to the macro data, remittances increased from about US$0.25 billion in 2003 to 1.4 billion in 2007, which is 40 percent of the country's GDP. By most conservative estimates the increase in remittances between 2003 and 2007 is accountable for at least 50 percent of the total observed poverty reduction.20 The simulations indicate that had remittances in 2007 remained at their 2003 level, the total poverty headcount would be around 63.2 percent, and the extreme poverty headcount would be around 27.9 percent. In other words, compared to the actually observed 2007 poverty level, the incidence of both total and extreme poverty would be about 10 percentage points higher (Figure 1.2). 20 If one uses macro data on the volume of remittances rather then the estimates obtained from the 2007 Household Budget Survey data (which is likely to underestimate remittances), the contribution of remittances to the total poverty reduction increases to as much as 90 percent. 17 Figure 1.2: What would be the Expected Poverty in 2007 had the Volume of Remittances Stayed at the 2003 Level? 80 total poverty 70 extreme poverty P vryha c u t % 72.4 60 oet edo n , 63.2 50 53.5 40 41.5 30 20 27.9 10 17.1 0 03' 07' 07* Note: 2003/2007­actual level of poverty in 2003/07. 07* - Simulated level of poverty in 2007 had remittances stayed at the 2003 level. Source: World Bank estimates using 2003 and 2007 TLSS data. The Expected Poverty Impact of the Financial Crisis The global financial crisis, which unfolded in late 2008, has adversely affected two of the most important destination countries for labor migrants from Tajikistan--Russia and Kazakhstan. As a real economic consequence of the crisis, demand for foreign labor is expected to fall in these countries. Migrants who have difficulties in finding further employment or who face serious wage cuts may consider returning, or even be forced to return, to Tajikistan. The resulting shortfall in remittance receipt will predominantly hurt those households that are highly dependent on remittances. According to the recent data on remittances flow reported by the National Bank of Tajikistan, remittances have already declined by 20 percent compared to the respective period in 2008. Using the TLSS 2007, we estimate the impact of reduced remittances receipt on diverse poverty and inequality measures in Tajikistan.21 We perform the simulations using two methodological approaches. The first approach (methodology 1) is simpler in that it simulates the impact on poverty and inequality of a certain reduction in remittances, assuming that at least in the short term the households would not be able to compensate for an induced loss in remittances. This approach also assumes a universal decline in remittances (by the same percent) across households. Using this approach, we simulate the poverty impact under the scenarios of a 20 percent, 30 percent, and 50 percent reduction in remittances. The second approach (methodology 2) is different in that we simulate the poverty impact of a certain percentage of migrants losing their jobs abroad (with a shock being randomly distributed across households). Hence, the assumption is that the main impact takes place through the loss of the jobs abroad, with those migrants who are able to retain their jobs abroad being paid the same monthly wage. In this approach, for return migrants (those who lost jobs abroad) we impute local monthly wages in the respective sector of employment (mostly construction) to account for the fact that in the longer term, return migrants will choose to work in the local labor market to compensate for the loss of incomes abroad.22 Hence, the second approach is based on a somewhat more optimistic scenario. Using this approach, we simulate the poverty impact under the scenarios of a 20 percent, 30 percent, and 50 percent loss of migrant employment, which is, however, partly compensated by taking jobs locally. Before presenting the results of the simulations, we provide a brief description of the nature and scope of Tajikistan's migration and highlight the country's exposure to (dependence on) remittances.23 21 The TLSS was fielded during October­November 2007, and hence represents a good baseline for measuring the impact of the crisis. 22 We also model the probability of being employed in the domestic labor market based on various individual characteristics. 23 Both methodologies effectively focus on measuring the direct impact of reduction in migration and remittances on poverty. However, in the longer term, there could also be complex knock-on effects. For instance, an increased inflow of return migrants 18 While remittances are a significant part of the story behind the 2003­07 economic growth and poverty reduction, they are also a source of vulnerability in times of the financial crisis. As discussed, the poverty reduction in Tajikistan over the last few years was largely driven by increasing migration and remittances. However, the same globalizing factor that has helped spur economic growth and poverty reduction is now transmitting the global shocks to the Tajik economy and households, due to the nature of Tajik migration. About 96 percent of migrants go to Russia, and of those, more than 51 percent chose Moscow as their destination, and 55 percent work in construction. Average monthly earnings abroad are US$300, three times the average earnings in Tajikistan (when computing the monthly earnings per working adult). High dependence on remittances manifests in the fact that 24 percent of all households have at least one migrant and that remittances have a significant bearing on household consumption.24 Simulations of the impact on poverty of a reduction in remittance flows to Tajikistan confirm the country's substantial external dependence and poverty impact.25 Using methodology 1, we find that remittances from abroad would halve in the course of the global financial crisis,26 and the national poverty headcount would rise from 53.1 percent to 60 percent (Table 1.10). The regional poverty headcount ratios would rise by between 2 percentage points (in urban Khatlon) to 13.6 percentage points (rural GBAO), reflecting various levels of regions' dependence on remittances. At the regional level, oblasts with a high share of households receiving remittances (width of remittance dependence) also exhibit higher shares of consumption financed through remittance receipt (depth of remittance dependence). Table 1.10: The Impact of the Crisis on the Poverty Headcount, under Various Scenarios of Reduction in Remittances % decline in Simulated poverty levels under Simulated poverty levels under remittances/employment methodology 1 (% decline in methodology 2 (% decline in abroad remittances) empoyment abroad) urban rural total urban rural total Current (baseline) 49.3 54.4 53.1 49.3 54.4 53.1 -20% 51.4 58.6 56.8 50.4 56.3 54.8 -30% 52.6 59.7 57.9 51.1 57.3 55.7 -50% 53.8 61.5 59.6 51.9 58.2 56.5 Note: The poverty headcount is based on the absolute national poverty line derived from TLSS 2007. (This poverty line is equal to 139 somoni per capita per month or, currently, US$1.4 per capita per day.) Source: World Bank estimates based on TLSS 2007. Using methodology 2, the poverty impact is somewhat lower, although still substantial. For instance, using methodology 1, a 30 percent decline in remittances would increase the poverty headcount from 53.1 percent to 57.9 percent, while using methodology 2, a 30 percent decline in employment abroad (with a substitution by local employment) would lead to an increase in the poverty headcount from 53.1 percent to 55.7 percent (Table 1.10). Rural areas are more affected than urban areas. The strongest poverty increase can be expected in the GBAO region, where the simulated (using methodology 2) rural headcount ratio would could push down wages in the domestic labor market, declining remittances could decrease domestic demand for goods and services, and so forth. 24 Those issues are discussed in detail in Chapter 4, on migration. 25 All effects are pure direct remittances effects, without taking into consideration the potential economic recession in Tajikistan (that is, worsening conditions in the local labor market) and without accounting for potential second-order effects from remittances (the multiplier effect). 26 Since the design of the migration module in the TLSS 07 underestimates remittances compared to the macro data (transfers captured by the banking system), a simulated decline of 50 percent at the household level corresponds to about a 25 percent decline at the macro level, which, according to the latest macro data, is what already seems to be happening. 19 rise by 9.5 percentage points and the urban headcount ratio would rise by 12.4 percentage points. Regions with lower migration incidence, like urban Khatlon, will be much less affected by the crisis (Annex 8, Table A8.1). The decline in remittances/employment abroad would also lead to an increase in the poverty depth (that is, those already poor will become poorer).27 The depth of poverty will increase by almost 50 percent on average, nationally (from 14.9 percent to 21.5 percent below the poverty line), while urban areas will experience 20 percent deeper poverty. The GBAO's recent success is especially endangered by the crisis-- the poverty gap in urban areas is expected to rise by a factor of three. Decline in remittances is also expected to increase inequality. In the scenario of the 30 percent decline in employment abroad, the Gini Coefficient increases from 32.1 percent to 35.1 percent28 (Table 1.10). As indicated by the strongly falling p10/p50 ratio,29 the main inequality effect stems from widening inequality in the lower tail of the welfare distribution. This is supported by evidence that most remittance-receiving households are highly dependent on this source of income. At the same time, reduction in remittances reaches all welfare quintiles: under the simulated scenario, 12 percent of households in the fourth quintile and 7 percent of households from the highest quintile would fall into poverty, suggesting that external financial dependence on remittances is high at all welfare levels. Table 1.11: Estimated Impact of the Financial Crisis on Consumption Inequality p90/p10 p90/p50 p10/p50 p75/p25 Gini Situation 2007 3.608 2.007 0.556 1.918 0.321 Simulation Minus 20% of Migrants (2007) 3.941 2.038 0.517 1.966 0.339 Simulation Minus 33% of Migrants (2007) 4.168 2.045 0.491 1.995 0.351 Simulation Minus 50% of Migrants (2007) 4.560 2.050 0.450 2.045 0.365 Note: Percentile ratios ignore households with zero consumption and are thus a lower bound estimate. Source: World Bank estimates based on the TLSS 2007. Conclusions The analysis indicates that Tajikistan achieved significant progress in poverty reduction between 2003 and 2007. The PPP US$2.15 per day poverty headcount declined from 64 percent in 2003 to 41 percent in 2007. This decline in the poverty headcount in terms of the absolute number of the poor translates into 1.35 million people (with 1 million of those in rural areas) escaping poverty, despite the increase in the total population by an estimated 0.4 million between 2003 and 2007 (to an estimated 7.1 million in 2007). Economic growth during 2003­07 has been pro-poor. The analysis indicates that lower deciles of the per capita consumption distribution have registered higher rates of growth. Indeed, the per capita consumption of the poorest 20 percent (1st quintile) of the population grew at an average annual rate of 13.6 percent, compared to the 8.3 percent for the 3rd (medium) quintile of the distribution, and 5.7 percent for the richest quintile. The household self-assessed changes in economic status also indicate that living standards of the population improved during 2003­07. For instance, 55 percent of the population lives in households where household heads believe that the financial situation improved over the last three years. Among the 27 It would also lead to a decline in the consumption of non-poor households (that is, increasing their vulnerability). 28 We obtain similar estimates using methodology 1. 29 This is the ratio of the per capita consumption of the poorest 10 percent of the distribution to the per capita consumption of 50 percent of the distribution (of all population). 20 households that are quantified as consumption-poor, this share is also quite high, at 48 percent. Only 14 percent of the poor population reported that their financial situation deteriorated over the last three years. The increase in migration and in related remittances is the most important factor behind poverty reduction during 2003­07. By most conservative estimates the increase in remittances between 2003 and 2007 accounts for at least 50 percent of the total observed poverty reduction. The simulations indicate that had remittances in 2007 remained at their 2003 level, the incidence of both total and extreme poverty would be about 10 percentage points higher compared to the actually observed 2007 poverty level. However, there is a real risk that poverty and inequality would be on the rise again in times of economic crisis. This is because migration and remittances are identified as the main transmission channels of the global financial crisis to Tajikistan. Using various methodological approaches (discussed in detail in Chapter 2 on poverty dynamics), we find that a 30 percent decline in remittances would increase the poverty headcount from 53.1 percent to 57.9 percent, while a 30 percent decline in employment abroad (with a substitution of lost employment abroad by local employment), would lead to an increase in the poverty headcount from 53.1 percent to 55.7 percent. Rural areas are expected to be more affected than urban areas due to a higher concentration of migrants there. The decline in remittances/employment abroad would also lead to an increase in poverty depth. In other words, those who are already poor will become poorer. The depth of poverty due to the impact of the crisis is estimated to increase by almost 50 percent on the national level (from 14.9 percent to 21.5 percent below the poverty line). A decline in remittances would also lead to a decline in consumption of non-poor households, hence increasing their vulnerability. The Government of Tajikistan is carefully monitoring the remittances situation and developing labor policy instruments that could absorb the potential increase in the supply of return migrants in the domestic labor market (such as public works or large infrastructure projects). A follow-up survey to measure the real impact of the crisis could provide much insight. 21 Chapter 2: Poverty Profile in Tajikistan30 A. Introduction This chapter presents a profile of poverty based on the 2007 Tajikistan Living Standard Survey (TLSS), which was fielded in October­November 2007. The TLSS 2007 was based on a random two-stage probability sample, stratified by oblast and type of settlements (urban/rural). The sample size was 4,860 households and it was representative at both the national and regional level. The TLSS is a multi-topic survey that allows linking several characteristics of households. The analysis focuses on various dimensions of the living standards. These include: (a) consumption- based poverty and inequality, (b) structure of incomes and income poverty, (c) self-assessed (subjective) poverty based on various indicators, and (d) food security and nutritional status. The TLSS is a part of a collaborative project of the State Statistical Committee of Tajikistan (GosKomStat), the World Bank, the United Nations Children's Fund (UNICEF), the Food and Agriculture Organization (FAO) of the United Nations, and the U.K. Department of International Development (DFID). B. Poverty and Inequality in Tajikistan This section presents a general picture of consumption-based poverty and inequality in Tajikistan using the TLSS 2007 data. The welfare measure used here is consumption per capita, and the poverty lines are the absolute total and extreme (food) lines (described in detail in Annex 3). Despite improvements in living standards over the last few years, poverty in Tajikistan remains widespread. At the end of 2007, 53.5 percent of the population was poor and 17.1 percent was extreme poor. In other words, every third poor was extremely (or food) poor. As a reflection of the higher level of extreme poverty in total poverty, the poverty gap is quite noticeable--the average consumption of the poor falls short of the overall poverty line by 15.4 percent (Table 2.1). Rural areas tend to be somewhat poorer than urban areas. Indeed, the poverty incidence in urban areas is only 5.6 percentage points lower than in rural areas--49.4 percent compared to 55 percent, respectively. As the analysis of the poverty trends indicates, the gap in the poverty incidence between urban and rural areas has narrowed since 2003 since the rate of poverty reduction was faster in rural areas, allowing them to catch up. When it comes to levels of extreme poverty, urban areas actually have somewhat higher rates--18.9 percent in urban areas compared to 16.4 percent in rural areas (Table 2.1). 30 This chapter was prepared by Oleksiy Ivaschenko (Economist, ECSHD), Sasun Tsirunyan (Consultant, UNICEF), and Yeva Gulnazaryan (Consultant, UNICEF). 22 Table 2.1: Consumption-based Poverty in Tajikistan Squared Headcount Poverty Poverty Rate(P0) Gap(P1) Gap(P2) Poverty Line = Total (139 Somoni) Urban 49.4 15.4 6.5 Standard Error 2.7 1.2 0.7 Rural 55.0 14.9 5.6 Standard Error 1.5 0.7 0.4 Total 53.5 15.0 5.8 Standard Error 1.3 0.6 0.3 Poverty Line = Extreme (89 Somoni) Urban 18.9 4.3 1.5 Standard Error 2.2 0.6 0.3 Rural 16.4 3.1 0.9 Standard Error 1.3 0.3 0.1 Total 17.1 3.4 1.0 Standard Error 1.1 0.3 0.1 Source: World Bank estimates using TLSS 2007 data. The analysis indicates that even a relatively modest change in the purchasing power of the population in Tajikistan will induce significant changes in the rates of poverty. A 10 percent decline in the purchasing power (approximated here by the 10 percent increase in the poverty line) would increase the poverty headcount from 53.5 percent to 62.6 percent, or by almost 10 percentage points. In the current environment of financial crisis and an expected reduction in remittances in Tajikistan, such a scenario is increasingly realistic. The good news for policymakers, though, especially in the longer (after crisis) perspective, is that this sensitivity generates the potential for further rapid poverty reduction given that the purchasing power of the population increases. Indeed, a 10 percent rise in purchasing power (approximated by a 10 percent decline in the poverty line) would reduce poverty from 53.5 percent to 44.5 percent (Table 2.2). Table 2.2: Sensitivity of Poverty Headcount to the Choice of Poverty Line Poverty Change from Incidence actual (%) (P0) Poverty Line = Total (139 Somoni) Actual 53.5 0.0 +5% 58.6 9.5 +10% 62.6 16.9 +20% 69.4 29.8 -5% 49.0 -8.4 -10% 44.5 -16.8 -20% 33.8 -36.9 Poverty Line = Extreme (89 Somoni) Actual 17.1 0.0 +5% 20.2 18.0 +10% 23.3 36.1 +20% 29.7 74.0 -5% 14.4 -15.7 -10% 12.0 -29.5 -20% 7.4 -56.9 Source: World Bank estimates using TLSS 2007 data. 23 To fully understand poverty, one has to keep in mind that poverty is a function of average consumption and consumption inequality. This distribution is characterized by the mean and dispersion around the mean (captured by the means for various groups, and various inequality measures). This representation is useful when one thinks about the poverty numbers. The mean per capita consumption in Tajikistan is equal to 157 somoni per month.31 It varies from an average of 148 somoni in rural areas to 180 somoni in urban areas.32 In other words, urban areas are on average 21 percent richer than rural areas. It also varies from the average of 74 somoni for the poorest per capita consumption quintile to the average of 302 somoni for the richest consumption quintile, implying that the richest 20 percent of the population spends on average 4.1 times the average of the poorest 20 percent of the population (Table 2.3). Table 2.3: Mean Consumption Per Capita for Different Groups Location Urban 180.2 Rural 148.6 Per capita consumption quintile Lowest quintile 74.0 2 106.4 3 133.2 4 169.8 Highest quintile 301.5 Total 156.9 Source: World Bank estimates using TLSS 2007 data. Inequality, measured by the Gini Coefficient, is relatively low. The Gini Coefficient of per capita expenditure is relatively low at 28.8 percent. This overall low inequality (under 30 percent) is driven mostly by low inequality in rural areas, where the Gini Coefficient is 25.4 percent. The inequality is much higher in urban areas, with a Gini Coefficient of 36 percent. The gap between the richest and the poorest is pronouncedly higher in urban areas (Table 2.4). The compressed distribution, especially in rural areas, also means that economic growth could bring about substantial poverty reduction, if it is broad based. Table 2.4: Inequality in Consumption Per Capita Distribution by Different Groups Bottom Half of the Upper Half of the Interquartile Tails Distribution Distribution Range p25/p10 p50/p25 p75/p50 p90/p50 p75/p25 p90/p10 Gini Total 1.33 1.33 1.36 1.84 1.81 3.24 28.8 Urban 1.34 1.43 1.46 2.13 2.09 4.09 36.0 Rural 1.31 1.30 1.33 1.74 1.73 2.97 25.4 Source: World Bank estimates using TLSS 2007 data. 31 This is very close to per capita GDP from the National Accounts (NA). 32 The differences by oblast (region) are discussed in detail later. 24 C. Poverty Profile and Multivariate Analysis of Poverty The TLSS 2007 contains extensive modules on various characteristics of households. These include place of residence, demographic composition, housing situation, access to facilities, employment sector of adult household members, migration, and education attainments. This section provides a profile of consumption-based poverty with respect to these key characteristics. C1. Geographic dimension of poverty A poverty profile describes the poor by indicating the risk of being poor according to various characteristics, such as type of employment, level of education of the household head, and demographic composition of a household (that is, gender, household size, number of children, ethnic status). Poverty in Tajikistan continues to be entrenched in rural areas. Of the overall population, 73.7 percent live in rural areas, and, as the result of the poverty incidence combined with these demographics, they account for 75.7 percent of all poor (and 70.9 percent of extreme poor) in Tajikistan (Table 2.5). Table 2.5: Poverty Headcount and Concentration, by Geographic Region Poverty Distribution Distribution Headcount of of the Poor Rate Population Poverty Line = Total (139 somoni) Urban 49.4 24.3 26.3 Rural 55.0 75.7 73.7 Oblast (region) Dushanbe 43.3 7.6 9.4 Sogd 68.8 38.1 29.7 Khatlon 47.3 31.5 35.7 RRS 48.8 20.2 22.2 GBAO 43.4 2.5 3.1 Total 53.5 100.0 100.0 Source: World Bank estimates using TLSS 2007 data. While at the national level the difference in poverty headcount between urban and rural areas is not that large, at the oblast level the difference is much more pronounced.33 For instance, in urban Sogd, the poverty incidence is on a par with the levels observed in urban areas of Khatlon and the Region of Republican Subordination (RRS). However, in rural Sogd, the poverty incidence is much higher than in urban Sogd. At the same time, the poverty incidence in rural Khatlon and RRS is actually lower than that observed in urban areas of those oblasts. In Gorno-Badakhshan Autonomous Oblast (GBAO), rural areas are much poorer than urban areas, which are effectively limited to the oblast center, Khorog (Figure 2.1). 33 This means that urban-rural differences at the oblast level get "averaged out" at the national level. 25 Figure 2.1: Urban Compared to Rural Differences in the Poverty Headcount, by Oblast 80% 74.0% 70% 56.8% 60% p v rty h a c u t edon 53.6% 52.5% 46.2% 47.6% 47.2% 50% 43.3% 40% 30% oe 18.4% 20% 10% 0% U -R e -U -R -U -R -U -R nb - n n O O ha P P d d tlo tlo og og R R A A us R R B B ha ha S S D G G K K Source: World Bank estimates using LSMS 2007 data. Altitude is highly correlated with poverty levels. In Tajikistan, many settlements are located at very high altitudes,34 where the weather conditions are extremely harsh and the main source of gainful employment comes from temporal migration (to Russia). The correlation between altitude and poverty can be clearly seen in GBAO, where many villages are located at more than 2 kilometers above sea level. Mostly as a result of those pronounced differences in altitude, the poverty headcounts across areas vary from 20 percent to 70 to 80 percent (Figure 2.2). Figure 2.2: The Correlation between Altitude (at the Primary Sampling Unit level) and Poverty in GBAO 90% 80% 70% poverty headcount 60% 50% 40% 30% 20% 10% 0% 0 500 1000 1500 2000 2500 3000 3500 4000 4500 altitude, meters Source: World Bank estimates using LSMS 2007 data. 34 The TLSS 07 measured global positioning system (GPS) coordinates and altitude at the cluster (Primary Sampling Unit) level. 26 C2. Employment and education of the household head The story of poverty in Tajikistan is largely that of the working poor.35 Employment, and especially non- agriculture wage employment, does reduce the risk of poverty. However, employment does not guarantee a poverty-free life. In fact, 58 percent of the poor come from households with employed household heads. Among various employment categories, non-agricultural wage employment and self-employment are correlated with the lowest risk of poverty (Figure 2.3). Employment in agriculture is not associated with a reduced risk of poverty. In fact, those households where the household head is wage-employed in agriculture face a higher risk of poverty than the national average--58.3 percent and 16.3 percent for total and extreme poverty, respectively. Self-employment in agriculture is also associated with a high risk of poverty (Figure 2.3). Self-employment outside of agriculture is associated with a somewhat lower risk of poverty than non- agriculture wage employment. Households with heads self-employed in non-agriculture have a total poverty headcount of 44.8 percent--compared to 47.8 percent in non-agriculture wage employment. However, they face a similar risk of extreme poverty--13.7 percent and 13.2 percent, respectively. Figure 2.3: Poverty by Household Head's Status of Employment Total povety, % 70 60.4 Extreme poverty, % 58.3 60 52.7 53.5 47.8 50 44.8 40 30 21.0 18.9 17.1 20 13.2 16.3 13.7 10 0 Not employed Wage employed Wage-employed Self -employed Self -employed Total (non-agr.) (agr.) (non-agr.) (agr.) Source: World Bank estimates using TLSS 2007 data. Due to unfavorable domestic labor market opportunities, many households opt to send their members to work abroad as a poverty-coping strategy.36 Indeed, about 25 percent of households report having either a return migrant over the last year or someone still working abroad at the time of the survey. Among households participating in the TLSS, 26.4 percent reported having more than one migrant. Migration (employment abroad) does reduce the risk of poverty. The effect is especially evident when one looks at the share of migrants in total household size, not simply the number of migrants. Households with a share of migrants in excess of 20 percent have total and extreme poverty headcounts of 42.6 percent and 13 percent, respectively, compared to 54.4 percent and 17.4 percent, respectively, for households with a share of migrants under (or equal to) 20 percent (Figure 2.4). 35 A detailed analysis of the linkages between the labor market characteristics of the individuals and poverty is presented in Chapter 4, on the labor market. Here we report only the key findings. 36 Migration is discussed in detail in Chapter 4 of this report. 27 Figure 2.4: Poverty by Household's Migration (work abroad) Status 60% Share of migrants < 20% 50% 54.4% P v rty h a c u t Share of migrants > 20% edo n 40% 42.6% 30% 20% oe 17.4% 10% 13.0% 0% total extreme Source: World Bank estimates using TLSS 2007 data. Households with more-educated household heads are less likely to be poor. Differences in educational attainment of heads of households are reflected in considerably different poverty rates. The risk of total poverty increases from 37 percent for households with a head with a higher (university) education to above 60 percent for households with a head with a basic37 (or below) level of education (Figure 2.5). Figure 2.5: Poverty by Household Head's Education Level 9.0 37.0 Higher (university) extreme poverty 11.9 48.7 total poverty Secondary technical 13.5 49.9 Secondary special 20.4 57.9 Secondary general 22.9 61.2 Basic 17.4 62.8 Primary 67.0 None 27.3 0 10 20 30 40 50 60 70 80 Source: World Bank estimates using TLSS 2007 data. C3. Demographics High fertility in Tajikistan continues to be a significant barrier to reducing poverty. Households with three or more children account for 53.2 percent of the total population and 62.2 percent of the total poor. This group of households faces a 62.6 percent risk of total poverty. The total poverty headcount declines to 32.9 percent for households with no children (Table 2.6). Higher poverty rates among larger households are related to higher dependency ratios. 37 "Basic" refers to completion of nine grades. 28 Table 2.6: Poverty, by Demographic Composition Poverty Distribution Distribution Headcount of the Poor of Population Rate Poverty Line = Total (139 Somoni) Number of children (<15) No children 32.9 6.6 10.8 1 43.7 12.0 14.8 2 48.2 19.2 21.3 3 or more children 62.6 62.2 53.2 Household size 1 6.4 0.1 0.4 2 19.6 0.4 1.2 3 21.0 1.3 3.2 4 34.7 4.5 6.9 5 44.0 11.1 13.5 6 51.8 16.2 16.7 7 or more 61.3 66.5 58.1 Total 53.5 100.0 100.0 Source: World Bank estimates using TLSS 2007 data. We also analyzed econometrically the behavior of various household parameters (beyond remittances) in terms of their role in poverty reduction between 2003 and 2007. We do this by looking at the comparable regression analysis coefficients, using the same model specification (run on the same independent household characteristics, and same dependent variable, which is the natural log of household consumption per capita) for both years. We find that some of the parameters are more responsible for the increase of consumption from 2003 to 2007 and, consequently, for the reduction of poverty, compared to others (Annex 9). We also find that the nature and gradient of the relationship between most of the other key household characteristics and level of consumption have not changed significantly over time. However, for some characteristics, the magnitude of their relationship with consumption has changed. For instance, in both urban and rural areas the positive correlation between being employed and household per capita consumption increased between 2003 and 2007. In rural areas, medium-size landownership (11 sotkas38 to 20 sotkas) becomes more positively correlated with per capita consumption, while the role of the larger-sized (20 or more sotkas) landownership somewhat declined. Interestingly, in rural areas the positive correlation between secondary and higher education and consumption became somewhat weaker between 2003 and 2007, which is most likely related to the increased role of (mostly low-skill) migration and remittances during that period (Annex 4). D. Household Income Poverty The preferable measure of household welfare for the TLSS analysis is consumption. However, we also investigate the relationship between household characteristics and poverty when we take household income as a welfare indicator of households. The TLSS 2007 contains extensive modules on total earnings from different income sources--wage employment, non-agricultural enterprises, agricultural enterprises, remittances, and other income. For each individual in the household above 15 years of age, information on time worked, sector of employment 38 A sotka is equal to 100 square meters. 29 (occupation and industry), and amount earned from each economic activity is collected. This enables the construction of a per capita income aggregate, which in turn can be used to estimate the incidence of poverty. Households in Tajikistan derive their incomes from three main sources. These include: (a) wages (including salaries from employment and earnings from self-employment and small businesses), (b) value of own food production, and (c) remittances from abroad. These sources of income account for 44 percent, 27 percent, and 17 percent, respectively, of the total disposable income. The composition of disposable income of households differs considerably across urban and rural areas, regions, and quintiles. As would be expected, rural households rely much more on home production--the value of home-produced products accounts for 30 percent of their total disposable income. Rural households also depend more on remittances, which account for 18 percent of their disposable income, while this share is 15 percent for urban households (Figure 2.6). Figure 2.6: Composition of Household Disposable Income, Percent Composition of households disposable income 3.5 1.6 10.0 3.6 RURAL 58.0 44.1 26.3 Wages Social Ass istance 4.9 3.8 1.7 5.7 Private transfers URBAN 78.1 24.2 20.6 Consum ption of ow n produced food NET Incom e from agriculture Incom e from w ork ing abroad 3.8 2.2 7.8 4.2 Other Incom e TOTAL 63.3 38.9 24.8 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Source: World Bank calculations based on TLSS 2007 data. Poor households depend heavily on consumption from own agricultural production. The main source of income for households in the two poorest income quintiles is production for self-consumption (55 percent). This means that these wages for those households were ranked as the second-most-important source, constituting 30 percent of total disposable income. For the households ranked in the top quintile, wages are the most dominant source of income, accounting for 47 percent of total disposable income. The income-based poverty estimates are higher than estimates of poverty based on consumption. This is consistent with the fact that in the countries with significant shares of informal income, consumption is a more reliable measure of household welfare.39 The income-based total poverty estimate is 66 percent (compared to 53 percent for the consumption-based estimate). The extreme poverty estimates based on income are about 25 percentage points higher than those based on per capita consumption (41 percent compared to 17 percent, respectively). Income-based estimates show smaller disparities across regions. In general, the regional patterns are broadly similar, but income-based estimates show smaller disparities across regions. The ranking of regions also changes by using disposable income per capita as a welfare measure. For example, while Sogd was 39 Estimates of income-based poverty are obtained by comparing the national poverty line expressed in prices of the fourth quarter of 2007 and per capita income for 2007 (instead of per capita consumption) expressed in the same price levels as the poverty line. 30 classified as the poorest region according to household consumption, the income-based poverty measures suggest that the poorest region is Khatlon (72 percent poor including 43 percent extremely poor population). E. Other Evidence of Living Standards While consumption is generally used as a core measure of household welfare, it is important to examine other non-consumption measures of well-being. This section examines the incidence of subjective poverty, food security, and nutritional status. E1. Subjective poverty (financial situation) Another way to estimate poverty is to analyze self-reported information about the households' financial situation and adequacy of consumption on the basis of answers to self-reported (subjective) assessment of well-being available in the TLSS 2007 data. This analysis reveals several interesting findings. The self-reported incidence of poverty, measured as a percentage of individuals living in households with an (self-reported) unsatisfactory financial situation, is very close to the incidence of consumption poverty. The share of individuals living in households who reported being unsatisfied with their financial situation is 56.9 percent. The respective shares are 65.5 percent for consumption poor, and 47 percent for consumption non-poor. For the bottom quintile of the per capita consumption distribution this share rises to 72.5 percent (Figure 2.7). Figure 2.7: Percent of Individuals Living in Households Unsatisfied with their Financial Situation 70% Total 60% Poor 65.5% Non-poor 50% 56.9% 40% 47.0% 30% 20% 10% 0% Total Poor Non-poor Source: World Bank estimates using TLSS 2007 data. The self-reported monthly per capita income needed to satisfy basic needs is very close to the consumption-based total poverty line derived from the survey. Indeed, when asked about the minimal monthly income needed to satisfy basic food and non-food needs, the households report the mean of 135 somoni, while the consumption-based poverty line is equal to 139 somoni per month. The consumption-poor report a lower self-estimated minimum (123 somoni) compared to the consumption non-poor (149 somoni). E2. Food security and nutritional status At the time of the survey, 24 percent of individuals were living in households with inadequate (self- assessed) food consumption. Among consumption-poor such share was 32 percent, and among consumption non-poor such share was 15 percent (Figure 2.8). Note that the share of food insecure of 24 percent is 31 somewhat higher than the share of 17 percent under the food poverty line derived from the survey. In the bottom quintile of consumption distribution this share rises to 44 percent.40 Figure 2.8: The Share of Self-reported Food-insecure, Percent of Individuals 35% 30% Total 32.0% Poor 25% Non-poor 20% 24.0% 15% 14.8% 10% 5% 0% Total Poor Non-poor Source: World Bank estimates using TLSS 2007 data. The data indicate that the level of food security varies considerably through the year. Seasonality affects all welfare groups, and most severe shortages of food are experienced during February­April. Indeed, the highest share of food insecurity occurs during the winter and early spring months. Among extreme poor, food sufficiency declines from about 55 percent in autumn to 25 percent during late winter-early spring (Figure 2.9). Figure 2.9: Seasonality in Food Security, by Welfare Group (percent of food secure) Food Sufficiency by Poverty status 100% 90% Ex treme Poor 80% 70% Poor but not 60% e x treme poor 50% Non poor 40% 30% Tota l 20% 10% 0% Ot- 6 o- 6 e- 6 Jn 7 e- 7 u- 7 Ag 6 e- 6 a- 7 Ar 7 My 7 J l- 7 Ag 7 c0 N v0 D c0 a -0 u -0 S p0 F b0 p-0 a -0 J n0 u0 u -0 Mr0 Source: World Bank estimates using TLSS 2007 data. The slump in food sufficiency due to seasonality is much deeper in rural areas. Interestingly, during the autumn months there is no gap in the level of food security between urban and rural areas, but then the level of food sufficiency falls much more sharply in rural areas (Figure 2.10). Clearly, one reason is that the share of poor and the size of the poverty gap are larger in rural areas. But access to markets, especially in the winter and early spring months, also plays a role. 40 This share of self-reported food insecurity refers to the October­November period, which is actually the most bountiful period of the year due to the harvest season. 32 Figure 2.10: Seasonality in Food Security, by Location (percent of food secure) Percentage of households reporting adequate food availability 100 90 80 70 60 50 40 30 Dushanbe 20 Other urban Rural 10 0 Aug-06 Sep-06 Oct-06 Nov-06 Dec-06 Jan-07 Feb-07 Mar-07 Apr-07 May-07 Jun-07 Jul-07 Aug-07 Source: World Bank estimates using TLSS 2007 data. Reducing the number of meals and/or altering the structure of the consumption basket are very common strategies for coping with food shortages. The share of households consuming less than three meals per day increases from 20 percent for non-poor households to 30 percent for poor households, to 42 percent for extreme poor households. When asked about types of food consumed yesterday, 49 percent of poor households compared to 73 percent of non-poor households reported eating fresh meat. The respective numbers for milk and milk products were 51 percent and 72 percent. Households resort to borrowing and more working hours to deal with food insecurity. Among households that report worrying about insufficient food, 60 percent reported borrowing money to buy food; 40 percent reported borrowing the food itself.41 Thirty-five percent report doing additional work in the domestic labor market, while 12 percent report having working migrants to help alleviate the problem. Finally, about 19 percent resort to asking relatives or friends for help. As a result of food deficiency, the poorest 20 percent of the population consumes less than 2,000 calories per person per day. Indeed, the poorest per capita consumption decile consumes only 1,550 calories per person per day. At the same time, the richest decile consumes almost 4,000 calories per person per day. As is well known from the nutrition literature, long-term food deprivation especially among children, negatively affects various human development indicators.42 41 Due to the possibility of multiple answers to this question, the categories are not self-excluding (for instance, the household can borrow money to buy food, and/or can borrow food directly), and thus various categories may sum up to more than 100 percent. 42 The nutritional status of children is discussed in detail in Chapter 7, on health. 33 F. Conclusion A significant reduction in poverty has taken place in Tajikistan since 2003, but poverty still remains widespread and deep. Using the absolute poverty line derived from TLSS 2007, we find that at the end of 2007, 53.5 percent of the population was poor, and 17.1 percent was extreme poor. In other words, every third poor person was extremely (or food) poor. The poor find themselves on average to be 15.4 percent below the total poverty line. There is a noticeable clustering of the population around the poverty line. The important policy implication of this is that even a relatively modest change in the purchasing power of the population in Tajikistan will induce significant changes in the prevalence of poverty. Indeed, a 10 percent decline in the purchasing power would increase the poverty headcount from 53.5 percent to 62.6 percent, or by almost 10 percentage points. Seventy-six percent of all poor (and 71 percent of extreme poor) live in rural areas in Tajikistan. The much higher concentration of the poor in rural areas is mostly due to demographics (70 percent of the population lives in rural areas). Indeed, the poverty incidence in rural areas is only 5.6 percentage points higher than in urban areas--55 percent compared to 49.4 percent, respectively. As the analysis of the poverty trends has indicated, the gap in the poverty incidence between urban and rural areas has narrowed since 2003, since the rate of poverty reduction was faster in rural areas, allowing them to catch up. The story of poverty in Tajikistan continues to be largely that of working poor. Employment, and especially non-agriculture wage employment, does reduce the risk of poverty. However, employment in the domestic labor market per se does not guarantee a poverty-free life. In fact, 58 percent of the poor come from households with employed household heads. Many households opt to send their members to work abroad as a poverty-coping strategy. About 25 percent of households report having either a return migrant over the last year or someone still working abroad at the time of the survey. Among households with at least one migrant, 26.4 percent actually report more than one migrant. Migration (employment abroad) does reduce the risk of poverty. Households with a share of migrants in excess of 20 percent have total and extreme poverty headcounts of 42.6 percent and 13 percent, respectively, compared to 54.4 percent and 17.4 percent, respectively, for households with a share of migrants under 20 percent. High fertility in Tajikistan continues to be one of the key correlates of high poverty risk. Households with three or more children account for 53.2 percent of the total population and 62.2 percent of the total poor. This group of households faces a 62.6 percent risk of total poverty and a 21.9 percent risk of extreme poverty. The total/extreme poverty risks decline to, respectively, 32.9 percent and 9.2 percent for households with no children. Households with seven or more household members account for 58.1 percent of the total population and 66.5 percent of the total poor. Higher poverty rates among larger households are related to higher dependency ratios. Poor households are twice as likely as non-poor households to face the risk of food insecurity. At the time of the household survey, 24 percent of individuals were living in households with inadequate (self- assessed) food consumption. Among consumption-poor households the share of food-insecure was 32 percent, while among consumption non-poor such share was 15 percent. The level of food security varies considerably through the year. The most severe shortages of food are experienced during February­April. As a result of food deficiency, the poorest 20 percent of the population consumes less than 2,000 calories per person per day. 34 Chapter 3: Tajikistan Domestic Labor Market43 A. Introduction Chapter 2 showed that poverty in Tajikistan is a complex and widespread phenomenon. Even though Tajikistan's economy grew at an impressive rate of 8.7 percent per year during 2000­07, rebounding after a period of political instability and economic decline, the country remains the poorest and one of the most fragile of the Commonwealth of Independent States (CIS) countries. This chapter assesses the current labor market situation in Tajikistan with an emphasis on the working poor.44 The analysis of the labor market of Tajikistan presented in this chapter is based on the latest Tajikistan Living Standards Survey (TLSS) (2007). With surging labor migration from Tajikistan, analysis of the labor market would be incomplete without discussing this issue. While this chapter explores linkages of the Tajikistan labor market with migration, the primary focus is on the domestic labor market. A detailed analysis of migration is presented in Chapter 4. B. General Labor Market Issues Tajikistan is a mountainous, landlocked, low-income country in Central Asia with a population of 7 million, of which more than three-fifths are in the 15­64 age (working-age) group. The working-age population in Tajikistan consists of 4.2 million people (Table 3.1), divided almost equally between those who are a part of the labor force (2.2 million) and those who are out of the labor force (2 million). Table 3.1: Main Labor Market Indicators in Tajikistan (in levels), 2007 Total Population 7,016,518 Total Population Age 15­64 4,215,165 Total Labor Force 2,171,008 Employed 1,965,231 Unemployed 205,777 ReturnMigrantsa 99,349 Out of Labor Force 2,043,653 Students 573,038 Housewives 1,053,628 Retired 132,616 a. Return migrants are a part of the domestic labor force if they are either employed or unemployed in Tajikistan. Source: World Bank estimates using TLSS 2007 data. Less than half of the working population is employed and one-tenth of economically active workers are unemployed. About 1.9 million workers were employed, that is, worked during the reference period or held a job from which they were temporarily absent for some reason. They constitute 47 percent of the working population. More than 205,000 workers in Tajikistan were unemployed, that is, they did not work during the reference period but actively looked for a job or did not actively look for a job because they were discouraged or worked seasonally. 43 This chapter was prepared by Maria Laura Sanchez Puerta (Economist, HDNSP), Sowmya Srinivasan (Consultant, HDNSP), and Oleksiy Ivaschenko (Economist, ECSHD). 44 Working poor are defined as those employed workers whose monthly per capita consumption is below the poverty line. 35 More than 50 percent of the economically inactive workers are housewives. There are close to 1.1 million housewives in Tajikistan. As we will see later in the chapter, these women are primarily concentrated in the 25­54 age group and in rural areas. Twenty-eight percent of the inactive working population consists of students and 7 percent consists of retired workers. Labor market participation in Tajikistan is much lower relative to other CIS countries. The labor force participation rate is extremely low, at 52 percent, especially when compared to other CIS countries (Table 3.2). This can be mostly attributed to the female labor force participation rate of 36 percent. Further, less than half of Tajikistan's working-age population reports being employed, whereas this rate is much higher for all CIS countries. There was a wide gap between the labor force participation (LFP) rates of men and women in 2007. The female LFP rate was more than 30 percentage points lower than the male LFP rate in Tajikistan. Further, this gap in participation rates is much more prominent in Tajikistan than in other CIS countries. Table 3.2: Main Labor Market Indicators ­ Comparison with other CIS Countries (in percentages) Azerbai Tajikistana Uzbekistan Armenia -jan Belarus Georgia Labor Force Participation (LFP) Rate 51.5a 64.8 53.3 66.7 57.8 61.8 Employment Rate 46.6 57.7 48.6 61.2 52.2 53.2 Unemployment Rate 9.5 -- -- -- -- -- Female LFP Rate 35.6 57.1 47.7 60.8 52.6 49.3 Male LFP Rate 69.5 72.7 60.3 73.2 64.0 76.4 Kazakhstan Kyrgyz Moldova Russia Turkmenistan Ukraine Republic Labor LFP Rate 69.7 64.5 60.2 60.6 66.7 56.1 Employment Rate 64.7 58.7 55.6 56.3 59.6 52.2 Unemployment Rate 7.8 -- -- -- -- -- Female LFP Rate 65.0 55.4 53.5 54.6 60.6 49.6 Male LFP Rate 74.9 74.3 67.8 67.9 73.2 64.0 -- = Not available. Note: All analysis has been carried out using household-level weights. Source: a. World Bank estimates using 2007 Tajikistan Living Standards Survey data. All other countries: World Development Indicators, World Bank (2006). Labor market outcomes are particularly discouraging for young workers (Table 3.3). Only 35 percent of people aged 15­24 participated in the labor market in 2007. The unemployment rate at 18 percent is much higher for young workers than for the rest of the working-age population. This rate is particularly high for young men (22 percent). Labor force participation and employment rates are around 45 percent lower for men aged 15­24 compared to their middle-aged counterparts. Only 28 percent of young women participated in the labor market in 2007 compared to 43 percent of young men.45 Better-educated workers have more favorable labor market outcomes. Unemployment rates for workers with secondary general education and basic or less education are high at 11 percent each, but decrease for those with technical and vocational education (8 percent) and higher and graduate education (5 percent). 45 This chapter discusses employment in the domestic labor market and does not cover seasonal/temporal migrants working abroad. 36 Seventy-three percent of the total working-age population lives in rural areas. Employment, unemployment, and participation rates are only marginally better in rural areas compared to urban areas. Further, breaking up the analysis by gender reveals a similar pattern. Rural men and women have better labor market outcomes in terms of higher participation and employment rates and lower unemployment rates than their urban counterparts (Table 3.3). However, incomes in rural areas are much lower. Table 3.3: Main Labor Market Indicators (in percentages), 2007a Share of Labor Force Working Participation Employmen Unemployme Population Rate t Rate nt Rate All 51.5 46.6 9.5 Gender Male 46.8 69.5 61.6 11.3 Female 53.2 35.6 33.4 6.3 Age 15­24 38.5 34.6 28.6 17.5 25­54 54.3 64.2 59.7 6.9 55­64 7.2 46.2 44.4 4.0 Educationb Basic or less 29.3 33.3 29.6 11.2 Secondary General 51.0 52.6 47.0 10.6 Technical and Vocational 10.9 74.7 69.1 7.5 Higher and Graduate 8.8 82.2 77.8 5.4 Location Rural 73.1 52.9 48.0 9.2 Urban 26.9 47.7 42.9 10.2 Region Dushanbe 9.5 45.8 41.5 9.3 Sogd 29.4 54.0 49.9 7.6 Khatlon 35.3 54.1 49.0 9.4 RRS 22.4 46.3 42.3 8.8 Gbao 3.4 53.4 36.7 31.2 Ethnicity Tajik 76.3 49.7 45.0 9.4 Uzbek 22.4 57.4 51.9 9.4 Russian 0.7 50.5 46.1 8.8 Kyrgyz Republic 0.4 64.3 46.8 27.2 Tatar 0.1 55.0 53.4 2.9 Turkmen 0.0 50.0 50.0 0.0 Other 0.3 54.0 52.9 1.9 Poverty Non-poor 50.0 51.0 46.8 8.4 Poor 50.0 52.0 46.5 10.6 Total 100.0 a. For labor market indicators in levels, see Annex 10, Table A10.2. b. Education categories have been recoded as follows: Basic or less (none, primary, basic); Secondary General (Secondary General); Technical and Vocational (Secondary Special, Secondary Technical); and Higher and Graduate (Higher Education, graduate school/aspirantura). Source: World Bank estimates using TLSS 2007 data. 37 C. Non-participation The overall non-participation rate in the domestic labor market in Tajikistan is high at 48 percent. Non-participants include all working-age individuals who are neither employed nor unemployed, that is, out of the labor force. In what follows, we discuss who is more likely not to be participating in the labor market. Among all age groups, the non-participation rate is highest for workers aged 15­24. The non-participation rate decreases with education for all age groups (Figure 3.1). Non-participation rises to 54 percent for workers aged 55­64 due to retirement age.46 Figure 3.1: Non-participation by Age Group and Education Level Basic or less 15-24 Secondary general Technical and 25-54 vocational Higher and graduate 55-64 0.0 20.0 40.0 60.0 80.0 Source: World Bank estimates using TLSS 2007 data. A wide gap exists between non-participation rates for men and women. Women account for more than two-thirds of all non-participants (1.44 million). There are striking differences between female and male non- participation rates among all age groups, increasing from 15 percentage points in the 15­24 age group to 49 percentage points in the 55­64 age group.47 Reasons for non-participation vary significantly by age and gender. Almost 90 percent of the young men who do not participate in the labor force report being students, whereas only 35 percent of women age 15­24 state this reason for non-participation (Table 3.4). A high 92 percent of non-participating women age 25­54 report being housewives (631,736), and close to 96 percent of women in the 55­64 age group report being housewives or in retirement (117,273). 46 The eligibility conditions for an old-age pension are 63 years of age and 25 years of work for men and 58 years of age and 20 years of work for women. 47 For the older group, the gap between women and men is driven by different old-age pension retirement ages. 38 Table 3.4: Reasons for Not Participating in Labor Force by Gender and Age (percent), 2007 Male Female All 15­24 25­54 55­64 15­24 25­54 55­64 Student/Pupil 29.7 88.3 15.1 1.3 35.3 1.3 0.1 Housewife 54.7 -- -- -- 61.3 91.9 28.8 In Retirement 6.9 0.5 11.9 66.3 0.1 2.2 67.0 Handicapped 3.8 3.5 30.5 16.0 1.2 2.7 3.4 In Military Service 0.1 0.5 0.7 -- -- -- -- Awaiting Recall by Employer 0.4 0.7 4.9 1.7 0.0 0.1 -- Do Not Want to Work 3.6 5.6 30.2 12.9 1.9 1.3 0.5 Other 0.8 0.8 6.8 1.8 0.3 0.6 0.3 Total 100 100 100 100 100 100 100 -- = Not available. Note: Men who claimed to be housewives were dropped from the sample. Source: World Bank estimates using TLSS 2007 data. Urban non-participation rates are higher than corresponding rural rates (52 percent compared to 47 percent). Further, this gap does not seem to narrow when considering non-participation rates net of those attending school. As far as women are concerned, family or household obligations weigh heavily on their decision to remain out of the labor force, and more so in rural areas (75 percent) than in urban areas (70 percent). D. Unemployment The unemployment rate in Tajikistan is 9.5 percent. The unemployed are defined as those who did not work during the past 14 days but actively looked for jobs in the past 30 days or did not actively look for jobs during the past 30 days either because they believe they have no chance of getting a job or that there are no jobs (that is, unemployment numbers include individuals discouraged in their job search) or they are waiting for the busy season to begin. For the composition of the unemployed by different demographic characteristics, see Annex 10, Table A10.2. Unemployment rates decrease with age and education. The unemployment rate is highest for young workers with secondary education, at 19 percent (Figure 3.2). For workers above age 25, the unemployment rate appears to decrease with education. Among those aged 55­64, the unemployment rate is highest for workers with basic and vocational education (5 percent). 39 Figure 3.2: Unemployment by Age Group and Education Level 15-24 Basic or less Secondary general 25-54 Technical and vocational 55-64 Higher and graduate 0.0 5.0 10.0 15.0 20.0 Source: World Bank estimates using TLSS 2007 data. Female, rural, and older workers are less likely to be unemployed compared to their counterparts. Controlling for other factors, females face a 6 percent lower risk of unemployment than males. Risk of unemployment is higher for urban workers than rural, but the effect is not large. Workers in age groups 25­ 54 and 55­64 face a lower risk of being unemployed compared to young workers (2 percent and 4 percent, respectively). Long-term (more than one year) unemployment is not a significant problem in Tajikistan, with an incidence of about 10 percent. Most unemployed workers have been looking for jobs for 12 months or less. The highest incidence of long-term unemployment is observed for those workers with basic education (16 percent) and those living in the Sogd region (15 percent). The risk of long-term unemployment is equal between genders. Fifty-four percent of the long-term unemployed are concentrated in rural areas and 42 percent live in the Sogd region (Table 3.5). 40 Table 3.5: Long-term Unemployment Duration and Shares, 2007 Average Duration Share in Incidence of in Months Long-term Long-term Median Mean Unemployment Unemployment All 18 19.2 100 9.8 Gender Male 18 17.6 73.4 9.8 Female 24 23.8 26.7 9.9 Age 15­24 18 17.0 35.7 8.7 25­54 18 20.5 64.3 11.0 55­64 -- -- -- -- Education Basic or less 18 21.0 27.1 15.6 Secondary general 18 17.8 40.6 8.2 Technical and Vocational 24 24.0 17.1 9.6 Higher and graduate 14 14.6 15.2 8.7 Location Rural 18 18.7 54.1 9.5 Urban 18 19.9 45.9 10.2 Region Dushanbe 24 22.8 12.1 4.7 Sogd 18 16.8 41.9 15.3 Khatlon 18 18.0 20.0 11.6 RRS 24 22.2 21.0 11.5 Gbao 24 24.0 5.0 3.8 -- = Not available. Source: World Bank estimates using TLSS 2007 data. E. Employment The employment rate in Tajikistan is low, at 47 percent. Employed individuals are those who worked during the last 14 days or had a job but were temporarily absent from work for some reason, or those who have already found a job that would start later. The unemployment rate decreases with education for workers of all age groups (Figure 3.2). Workers with higher and graduate education have the highest employment rates at 56 percent in the 15­24 age group, 81 percent in the 25­54 age group, and 72 percent in the 55­64 age group. Agriculture and related activities dominate the employment sector in Tajikistan. Forty-five percent of the employment sector in Tajikistan is comprised of agriculture, fishing, and forestry-related activities (but mostly agriculture). The next-most-important sectors of employment are services and commerce, together constituting 30 percent of total employment. However, the sector structure of employment varies considerably by location and region. Fifty-seven percent of all rural employment is in the agriculture sector (Table 3.6). On the other hand, urban areas hold the largest concentrations of those working in the commerce and service sectors (55 percent). This pattern is similar in the capital city of Dushanbe (64 percent). 41 Table 3.6: Sectors of Employment by Location and Region (percent), 2007 Urban Rural Dushanbe Soghd Khatlon RRS Gbao Agriculture, Fishing, and Forestry 9.8 56.8 0.6 44.1 57.4 43.5 40.5 Mining and Quarrying 0.5 0.3 0.0 0.7 0.0 0.6 0.0 Manufacturing 9.1 4.3 7.6 6.4 3.1 8.3 0.8 Electricity, Gas, Water and Sanitary Services 2.8 1.1 2.3 1.0 1.9 1.3 1.1 Construction 8.6 8.4 9.4 6.5 7.1 14.2 2.7 Commerce, Research and Development 27.2 9.0 32.0 14.1 11.1 10.0 6.9 Transport, Storage, and Communication 8.6 4.0 12.4 4.6 3.3 6.1 5.8 Services 28.1 12.1 32.2 15.1 13.3 12.9 39.6 Other Activities 5.3 4.1 3.6 7.4 2.8 3.1 2.5 Total 100 100 100 100 100 100 100 Source: World Bank estimates using TLSS 2007 data. Piecemeal (no regular contract) employees and unpaid family workers constitute a substantial share of employed. Regular employees constitute only 40 percent of the employed population in Tajikistan. Piecemeal employees, who are paid compensation on a piecework basis, form the second-largest component (25 percent). Unpaid family workers account for 18 percent of the employed population (Table 3.7). Table 3.7: Types of Employment by Age, Gender, and Location, 2007 Age All Group Male Female 15­24 25­54 55­64 Rural Urban Rural Urban Employee 39.9 31.9 41.0 55.8 34.2 45.2 40.4 57.9 Piecemeal Employee 24.5 27.8 24.3 16.2 31.4 23.4 18.6 10.9 Self-employed Agriculture 4.8 3.5 5.1 7.3 6.4 0.9 5.6 1.2 Self-employed Non- agriculture 6.5 4.8 7.4 3.6 5.1 14.5 2.3 13.4 Employer 6.4 7.9 6.1 3.5 7.5 6.9 5.4 3.1 Unpaid Family Worker 17.5 23.6 15.9 13.2 15.1 8.6 27.4 12.8 Other 0.4 0.6 0.3 0.4 0.3 0.5 0.4 0.7 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Source: World Bank estimates using TLSS 2007 data. Compared to urban areas, rural areas hold larger concentrations of piecemeal employees, self- employed workers in agriculture, and unpaid family workers. These differences are more pronounced for women than for men for all employment categories except self-employed workers in agriculture. Better- educated workers are more likely to hold regular jobs. There are significant differences across sectors in the dominant type of employment. Employees are concentrated primarily in the agriculture (34 percent) and services (37 percent) sectors. Forty-seven percent of piecemeal employees work in agriculture, 22 percent in construction, and 14 percent in commerce. Almost 60 percent of self-employed outside agriculture work in commerce, and 17 percent work in the transport, storage, and communication sector. More than 80 percent of unpaid family workers work in the agriculture sector (Table 3.8). 42 Table 3.8: Composition of Employed by Sector of Employment and Type of Employment Piece- Self- Self- Employee employee employed employed Employer Unpaid Other Total Family Agriculture Non-agriculture Worker Agriculture, Fishing, and Forestry 266,178 192,041 93,876 0 46,576 274,476 3,154 876,301 Mining and Quarrying 3,919 2,126 0 876 0 209 0 7,130 Manufacturing 66,256 16,670 0 9,966 3,909 10,112 0 106,913 Electricity, Gas, Water and Sanitary Services 26,341 1,864 0 740 479 0 0 29,424 Construction 34,717 105,072 0 6,809 13,619 2,579 491 163,287 Commerce, Research and Development 48,492 65,812 0 72,706 34,748 37,838 1,600 261,196 Transport, Storage, and Communication 36,003 17,669 0 21,525 12,090 10,957 888 99,132 Services 283,391 19,402 0 6,020 1,360 1,143 1,169 312,485 Other Activities 9,423 55,751 0 7,773 10,756 1,592 0 85,295 Total 774,720 476,407 93,876 126,415 123,537 338,906 7,302 1,941,163 Source: World Bank estimates using TLSS 2007 data. More than one-third of workers are in the informal sector. Here, informality is defined as the lack of a signed contract with the employer in the primary job. By this definition, informality represents a large part of regular wage employment, with 36 percent of employees working without a contract. The informal part of the economy is dominated by the agriculture and services sectors. Forty-seven percent of employees working informally are in the agriculture, fishing, and forestry sector and 25 percent are in the services sector. At 8 percent, commerce holds the next-largest share of employees in the informal sector. The probability of working informally is significantly correlated with being female, less educated, and young. F. Labor Incomes and the Working Poor Labor incomes are lowest in the agriculture and services sectors (even after accounting for income from agriculture and value of in-kind consumption). At 80 somoni a month, median incomes in the agriculture sector are the lowest among all sectors of employment (Figure 3.3). The services sector has the next-lowest median monthly income at 159 somoni. The construction sector earns the highest median monthly income at 366 somoni followed by commerce (300 somoni) and transport (300 somoni). 43 Figure 3.3: Median Monthly Income by Sector of Employment 400 366 350 300 300 300 284 250 234 219 201 200 159 150 100 80 50 0 Agriculture, Mining and Manufacturing Electricity, Gas, Construction Commerce, Transport, services Other activities fishing and quarrying Water and storage and forestry Sanitary communication Services Source: World Bank estimates using TLSS 2007 data. Workers self-employed in agriculture earn the lowest income and those self-employed outside agriculture earn the highest. Median monthly incomes of workers self-employed in agriculture (taking into account income from agriculture and value of in-kind consumption) constitute less than 30 percent of median incomes outside agriculture (Figure 3.4). Further, unpaid family workers who constitute one-third of agricultural employment earn the lowest median incomes (54 somoni) among all employment types. Employees, who constitute close to 30 percent of employment in agriculture, earn median incomes that are half of the median incomes of workers self-employed outside of agriculture. Figure 3.4: Median Monthly Income by Type of Employment 350 300 300 250 201 200 173 151 142 150 100 54 54 50 0 Employee Piece- Self- Self- Employer Unpaid Other employee employed- employed- family agri nonagri worker Source: World Bank estimates using TLSS 2007 data. Labor incomes are lower in the public sector and state-owned enterprises compared to those in the private sector. Median monthly incomes of employees in the public sector (172 somoni) and state-owned enterprises (139 somoni) are lower than those in the private sector (178 somoni). 44 Incomes in the informal sector are lower than those in the formal sector. The unadjusted mean monthly incomes for formal sector employees are 23 percent higher than for informal employees. In rural areas, formal sector employees earn 28 percent higher incomes than informal sector employees, whereas in urban areas this difference is only 7 percent. The formal sector wage premium is 26 percent in agriculture and 12 percent in the services sector. The conditional regressions of log monthly incomes on the formality indicator and several individual characteristics indicate that formal sector earnings are at least 10 percent higher than informal sector earnings. Considering the level of labor incomes, it is not surprising that 50 percent of the employed are poor. Working poor are defined as those employed workers whose monthly per capita consumption (adjusted with strata price deflators) is below the poverty line. As discussed, the majority of all employment in Tajikistan is concentrated in the agriculture and services sectors, which are marked by the lowest incomes among all employment sectors. Hence, it is not surprising that half of the employed are living below the poverty line. The working-poor population is mainly composed of men, middle-aged workers, those in rural areas, poorly educated workers, and those working in the agriculture sector. G. Conclusions Tajikistan has a total population of 7 million and a working-age population of 4.2 million. The working- age population is further divided almost equally between those who are a part of the labor force (2.2 million) and those who are out of the labor force (2 million). There are 1.95 million employed, which constitutes 47 percent of the working-age population. There are 0.25 million unemployed, which puts the unemployment rate at 9.5 percent. Labor force participation in Tajikistan is much lower than the Europe and Central Asia (ECA) country averages. The labor force participation rate is extremely low at 52 percent, especially when compared to other ECA countries. This can be mostly attributed to the low female labor force participation rate of 36 percent, which is 30 percentage points lower than the male labor force participation rate in Tajikistan. More than 50 percent of the economically inactive workers are housewives. There are close to 1.1 million housewives in Tajikistan. These women are primarily concentrated in the 25­54 age group and in rural areas. Twenty-eight percent of the inactive working-age population are students and 7 percent are retirees. Education levels and non-participation rates are negatively correlated; that is, non-participation decreases with education for all age groups and across genders. Agriculture dominates the employment sector in Tajikistan. Agriculture accounts for 45 percent of the employment sector in Tajikistan. The next-most-important sectors of employment are services and commerce, constituting 30 percent of total employment. The employment rate increases with education for workers of all age groups. The key feature of the Tajik domestic labor market is the very high share of working poor. In fact, 50 percent of the employed are poor. Almost 80 percent of the working poor live in rural areas. Only 6 percent of the working poor live in the capital city of Dushanbe, whereas 75 percent live in the mostly rural regions of Sogd and Khatlon. Half of the working poor are in agriculture, with another 13 percent in the commerce and services sectors each. Low labor incomes and a high prevalence of temporary work arrangements, informality (no labor contract), and unpaid work are the main reasons there are so many working poor. Labor incomes are lowest in the agriculture and services sectors, even after accounting for income from agriculture and value of in-kind consumption. At 80 somoni a month, median income in the agriculture sector is below the food poverty line. The services sector has the next-lowest median monthly 45 income at 159 somoni. The construction sector has the highest median monthly income at 366 somoni, followed by the commerce and transport sectors. Rural areas hold larger concentrations of piecemeal (temporary work arrangement) employees, self- employed, and unpaid family workers. More than 80 percent of unpaid family workers work in the agriculture sector. Forty percent of piecemeal employees work in agriculture, 22 percent in construction, and 14 percent in the commerce sector. Almost 60 percent of self-employed outside agriculture work in commerce and 17 percent in the transport, storage, and communication sector. The prevalence of temporary work arrangements is more pronounced for women than for men across all sectors. More than one-third of employees are in the informal sector. The informal part of the economy is dominated by the agriculture and services sectors. Forty-seven percent of employees working informally are in the agriculture sector, and 25 percent are in the services sector. At 8 percent, the commerce sector holds the next-largest share of employees in the informal sector. The income premium for working in the formal sector is 10 percent once other determinants of income are taken into account. The overall unemployment rate is 9.5 percent, with variations by age and education. The unemployment rate is highest for young workers with general secondary education, at 19 percent. For workers over age 25, the unemployment rate decreases with education. Controlling for other factors, females face a 6 percent lower risk of unemployment than males. The risk of unemployment is higher for urban workers than for rural workers, but not by much. Long-term (more than one year) unemployment is not a significant problem in Tajikistan. Its share in overall unemployment is 10 percent. The highest incidence of long-term unemployment is observed for workers with basic education (16 percent) and those living in the Sogd region (15 percent). The pool of long- term unemployed is composed primarily of men, those in the middle-age group, poorly educated workers, and those in the Sogd region and rural areas. 46 Chapter 4: Labor Migration, Remittances, and Welfare Implications for Tajikistan48 A. Introduction Migration has long been perceived as an important tool for poverty reduction in Tajikistan despite the lack of clear empirical evidence. The few migration and remittances studies on Tajikistan are mostly based on small surveys or qualitative research. The aim of this chapter is to use the nationally representative Tajikistan Living Standards Survey (TLSS) 2007 data to investigate the following issues: (a) internal and external migration patterns in Tajikistan, (b) the profile of migrants and their earnings abroad, and (c) the impact of remittances on household welfare. The TLSS 2007 covers both return migrants, who are surveyed about their migration experience abroad, and currently absent household members, about whom information is provided by household members. Unfortunately, the survey requested only the last migration spell of returnees instead of full migration histories, which naturally introduces a bias in estimates of migration flows for recent years. The multipurpose TLSS allows one to investigate patterns and characteristics of migration spells, labor market integration of return migrants, and the receipt and use of remittances at the household level.49 B. Internal and International Migration Flows in Tajikistan B1. Internal migration Internal migration incidence rose drastically in the post-Soviet era. On a regional basis, internal migration patterns have changed between the civil war (1992­98) and postwar period in most oblasts. One finding is that at least 40 percent of migrants from all oblasts (except Sogd) have moved to Dushanbe, which clearly points to the important role of the capital city. Two-thirds of all migrants from the Region of Republican Subordination (RRS) oblast, which was greatly affected by the war, headed for the capital city of Dushanbe. Work- and study-related reasons have grown in importance for internal migration, while family reunion was an especially important migration factor in the early years of independence and during the war. Motivations for relocating differ markedly between the two periods, 1991­98 and 1999­2007 (Table 4.1). Security was an important reason for migrating from the most severely war-affected oblasts, Khatlon and RRS. At a later stage, the security situation in Dushanbe induced migration to (other) Tajik oblasts. 48 This chapter was prepared by Alexander Danzer (Consultant, ECSHD) and Oleksiy Ivaschenko (Economist, ECSHD). 49 While this chapter considers both internal and external migration, more emphasis is put on the analysis of external migration due to its significant role in economic development. 47 Table 4.1: Migration Reasons According to Source Region and Time Period Government Other Work Marriage Family Health Study Security Resettlement Reason (%) (%) (%) (%) (%) (%) (%) (%) Dushanbe 4.2 39.6 49.8 0.0 0.0 1.7 1.9 2.8 Sogd 0.7 19.9 70.6 0.0 0.0 0.0 8.8 0.0 Khatlon 2.9 23.9 65.9 0.2 0.6 4.8 0.9 0.7 1991­ RRS 6.3 36.1 46.8 1.1 1.5 3.0 2.0 3.3 1998 GBAO 10.6 33.3 49.8 0.0 0.0 0.0 0.0 6.3 Abroad* 8.9 34.9% 49.9 0.0 0.0 4.0 0.0 2.3 Total 4.0 29.3 58.7 0.4 0.6 3.4 1.9 1.7 Dushanbe 1.2 54.0 38.4 1.6 0.5 2.6 0.9 0.9 Sogd 6.7 12.5 69.0 0.0 5.7 0.0 0.0 6.2 Khatlon 7.9 53.2 30.8 0.3 2.0 0.0 4.0 1.8 1999­ 2007 RRS 5.3 49.1 41.6 0.3 1.2 0.0 1.7 0.9 GBAO 1.3 60.9 25.7 0.0 12.1 0.0 0.0 0.0 Abroad* 7.7 56.0 19.7 0.0 1.8 10.9 0.0 3.9 Total 5.8 48.9 37.9 0.5 2.1 0.6 2.4 1.8 * Abroad refers to persons resettling from foreign countries. Source: World Bank estimates using TLSS 2007 data. B2. Extent and types of international migration in Tajikistan since 2004 From January to October 2007, 350,000 people, or 5 percent of the entire population, was (or continue to be) abroad from Tajikistan for at least one month (Table 4.2). Between January and October 2007, around 100,000 migrants returned to Tajikistan. Given that virtually all international migrants are working age and that the share of the working population (either employed, self-employed, or farming) in the working-age population is relatively small in Tajikistan, the number of migrants in 2007 equaled almost 20 percent of the current working population.50 50 Unfortunately, the TLSS 2007 provides only limited information on past migration histories and only collects information on the last migration spell. A person having multiple short stays abroad will be counted only once. Furthermore, the TLSS cannot account for migration of single persons or the migration of entire households that cannot be sampled. Thus, we are certainly underestimating past migration flows. 48 Table 4.2: Total Migration in Tajikistan (TLSS 2007) Year Absolute Percent of Total Percent of Percent of Number in Population Working-age Working Thousands (%) Population Population (%) (%) Population (exclusive migrants) 2007 7,000 100 Working-age Population 2007 4,253 60.8 100 Working Population 2007 1,827 26.1 43.0 100 Estimated Return Migrants 2007 100 5.7 2.4 5.5 Estimated Current Migration 2007 250 9.2 5.9 13.7 Stock Total migrants 2007 350 5.0 8.2 19.2 2006 271 3.9 6.4 14.8 2005 143 2.0 3.4 7.8 2004 79 1.1 1.9 4.3 Note: Numbers for 2007 cover January­October 2007 only. Consequently, real absolute migration numbers are likely to be higher and the share of return migrants in total migrants might be downward biased. Total population and working-age population include return migrants. The numbers for return migrants refers to the last migration spell reported. Unfortunately, earlier years are likely to be severely underestimated. Source: World Bank estimates using TLSS 2007 data. B2.1 Destination choice Russia is the main magnet for Tajik migrants, attracting 96 percent of migrants, with more than 51 percent of those migrants heading for Moscow (Figure 4.1). Meanwhile, Kazakhstan has become a leading regional importer of labor since it faces labor shortages in many segments of its labor market. Over 200,000 migrants from other Central Asian states work in Kazakhstan. The share of migrants working in Kazakhstan is only 1.5 percent of the total number of Tajik migrants. Figure 4.1: Destination Choice of Tajik Migrants, by Country and Major Cities Other non CIS Moscow 3.5% 51.4% Other CIS Russia 0.3% 96% Sankt-Peterburg Kazakhstan 6.6% 1.5% Ekatarenburg 5.4% Other Russian city 27.8% Tyumen' 3.5% Source: World Bank estimates using TLSS 2007 data. Despite uniformity in the destination choice, reasons for specific destination choices are quite diverse. The destination choices are mostly driven by personal contacts (35 percent and 41 percent for 49 men and women, respectively) or prior migration experience in the destination region (23 percent and 26 percent for men and women, respectively). Especially striking is the difference in importance of household and family reasons and the difference in prearranged jobs between genders.51 While family reasons play a significant role for women (23 percent), prearranged jobs (28 percent) are especially important for men. B2.2 Seasonality of migration flows and migration duration The flows of return migrants show a strong upward trend, with clear seasonality patterns. In terms of the time of the start of the migration spell, there is an early peak during the spring and a smaller peak during harvest time (around September). Since 2006, the monthly emigration numbers increased markedly, reaching almost 8,000 in mid-2007 (Figure 4.2). Figure 4.2: Monthly Flow of Return Migrants, by Month of Exit (thousands) Source: World Bank estimates using TLSS 2007 data. About 95 percent of Tajik migrants are men. Women migrate, on average, for much longer periods than men, with their last migration spell lasting over 60 percent longer than that of male migrants. Although women are less likely to migrate abroad, they tend to stay abroad substantially longer, with almost half of female migrants returning after only 10 months, while less than one-third of male migrants stay that long. There is also a tendency toward longer migration spells for immigrants from less- accessible regions (for example, Gorno-Badakhshan Autonomous Oblast [GBAO]). B2.3 Profiles of migrants The typical migrant from Tajikistan is a male around 30 years old. Migrants are over- proportionally drawn from rural settlements and higher settlement altitudes, reflecting fewer options for domestic income diversification in these regions. Table 4.3 shows profiles of three types of migrants: those who went abroad and have not returned yet, those who returned after being away for more than a year, and those who returned after less than one year. There are differences among these groups as regards individual and household characteristics and geographic background. Migrants who have not 51 Unfortunately, the TLSS 2007 survey does not provide information on repeated migration. 50 returned yet are substantially younger and came especially often from GBAO oblast. Their respective households are smaller and less likely to be poor. Those return migrants who stayed away for more than one year are most often from Dushanbe and Kathlon and are less likely to go to Russia. Long-term migrants also earn much more than migrants with shorter migration experience. Table 4.3: Profiles of Different Types of Migrants in Tajikistan (I) (II) (III) Migrant Return Return T-tests Not Migrant away Migrant Away Returned More than One Less than One Yet Year Year Female 6.4% 11.6% 6.4% ** ## Age 28.4 34.9 34.7 ** §§ Age at Emigration 26.0 32.5 33.5 ** §§ Migration Duration, in Months 32.5 25.0 6.4 ** ## §§ Destination : Russia 95.3% 92.1% 95.8% * ## Dushanbe 10.0% 24.0% 15.6% ** ## §§ Khatlon 19.8% 30.2% 27.9% ** §§ Sogd 22.0% 15.5% 23.7% ** ## RRS 22.0% 16.3% 24.7% * ## GBAO 26.3% 14.0% 8.1% ** ## §§ Net Domestic Monthly Pay (somoni) n.a. 452.1 352.7 n.a. ## n.a. Household Size 6.2 7.6 7.2 ** # §§ Dependency Ratio in Household 33.3% 35.8% 35.0% § Employment Ratio in Household 27.9% 26.0% 26.9% Poor Household 40.4% 47.3% 50.9% * §§ n.a. = Not applicable. With respect to employment after returning to Tajikistan, if applicable. T-test statistics: ** means the difference is significant at 5 percent level and * means 10 percent level between (I) and (II). ## means difference is significant at 5 percent level and # means difference is significant at 10 percent level between (II) and (III). §§ means difference is significant at the 5 percent level, and § means difference is significant at the 10 percent level between (I) and (III). Source: World Bank estimates using TLSS 2007 data. Migrants with general secondary education are the dominant group. Less than every seventh migrant has lower education and a similar number hold a university degree. Those with higher education are definitely underrepresented in the migration process. Less than 60 percent of Tajik migrants work legally (that is, have a work permit). However, those who do work legally earn 10.5 percent more abroad when controlling for individual and household characteristics and regional fixed effects.52 Simple mean comparison analysis reveals that legal migrants are more likely to be female, younger, come from Sogd oblast, and live in smaller and less-poor households. The main reasons for returning stated by return migrants are homesickness and family-related reasons (46 percent and 53 percent of answers for men and women, respectively). The share of migrants 52 The regression results are not shown here for brevity. 51 returning because of seasonal work relationships is low--only 6.8 percent of men and 2.5 percent of women report seasonal work as the main reason for returning home (Table 4.4). Table 4.4: Return Reasons, by Gender Men (%) Women (%) No Residence Permit 5.5 3.8 No Work Permit 5.2 7.2 Permit Expired 3.5 5.6 No Intention 7.4 1.2 Enough Money 6.2 9.5 Seasonal Work 6.8 2.5 Expelled 2.9 2.9 Family Reasons 18.3 20.5 Homesick 27.7 32.4 Other 16.5 14.4 Total 100 100 Source: World Bank estimates using TLSS 2007 data. B2.4 Labor market relations and remittance transfers The majority of migrants work in the construction sector (55 percent) or perform other unskilled tasks (22 percent). Monthly incomes generated by migrants abroad are especially high in professional occupations, sales, and housekeeping. Those are also the sectors with the highest rates of legal jobs (Table 4.5). Current migrants are generally similar in their profile to return migrants and mostly differ with respect to duration of their stay abroad. Table 4.5: Migration Characteristics of Return Migrants by Occupation Abroad Average Net Income in US$ Employment Legal Work (%) (monthly) Months Away Share (%) Construction 52.2 323.9 9.1 54.5 Unskilled 45.2 270.6 8.5 17.6 Professional 86.5 405.7 10.3 4.8 Driver 76.6 385.8 14.6 3.5 Sale 61.4 308.5 11.0 8.9 Agriculture 84.9 187.1 14.1 3.9 Housekeeping 100.0 332.5 9.6 1.4 Other occupation 70.8 301.4 11.0 5.4 Total 57.8 313.4 9.7 100 Note: Table comprises only migrants with full information. Source: World Bank estimates based on the TLSS 2007 data. In all occupations but agriculture, migrants earn abroad on average six times the monthly average income of workers in Tajikistan. Drivers and professionals abroad earn around eight times the average Tajik monthly wage, while employment in agriculture earns only about four times the average income at home. However, the average monthly labor income in agriculture in Tajikistan is only US$25 per month and unpaid subsistence farming is widespread. Thus, migration seems very attractive for workers in this sector. 52 Migrants' pay abroad is positively associated with age and education. If someone has a university education, they still earn only 7 percent more than someone with a secondary degree and 20 percent more than someone with lower education or no degree. Due to their better domestic labor market prospects, university graduates are less likely to go abroad. The large majority (74 percent) of currently absent household members send monetary remittances. While some send remittances in the form of cash and goods (6 percent), the transfer of only goods in-kind is rather uncommon (1 percent), which underpins the role of remittances for relaxing the cash constraint of households. However, more than 18 percent send no remittances to their household in Tajikistan at all. B2.6 Labor market reintegration in Tajikistan A substantially smaller share of return migrants works (being employed, self-employed, or working on a farm) domestically compared to non-migrants (Table 4.6). The analysis indicates that it is unlikely that return migrants are not employed in the domestic labor market because they await the next migration spell. However, returned migrants currently achieve much higher earnings compared to those who have never been abroad. This earnings advantage, which can be attributed either to skill improvements abroad or to sample selection, shrinks to a statistically significant 28 percent after running a multivariate regression controlling for age, gender, marital status, employment type, and oblast dummies. Table 4.6: Labor Market Success of Return Migrants in Tajikistan (aged 16­65) Men Women** Non- Migrant Diff. Non- Migrant Diff. migrant migrant Worked (employed, self-employed, or farming) 58.5% 50.0% ### 33.1% 23.2% Employed* 33.6% 25.1% ### 17.5% 18.8% Self-employed* 13.5% 14.6% 4.1% 5.3% Farming* 14.7% 11.8% # 13.6% 1.7% ## Hours worked/week 46.3 45.6 42.0 44.1 Net pay per month (somoni) 346.7 460.0 ### 171.2 644.4 ## *Categories add up to higher shares than "worked" due to multiple responses. ** Values for migrant women have to be interpreted with caution because the sample size is small. ### indicates statistical significance at the 1 percent level, ## indicates statistical significance at the 5 percent level, and # indicates statistical significance at the 10 percent level. Source: World Bank estimates using TLSS 2007 data. C. The Welfare Implications of Remittances for Households Poorer households are much more likely to send household members to work abroad. While 60 percent of households in the lowest pre-remittance consumption quintile have now or had in the past household members abroad, the share of better-off households with migrants is 13 percent (Table 4.7). However, since households base their domestic labor supply and income generation decisions on the given receipt of remittances, it is difficult to infer what could have been the true level of welfare (that is, the counterfactual) in the absence of remittances. 53 Table 4.7: Share of Households with Migrants, by Pre-remittance Consumption Quintiles Pre-remittance Number of Return Migrant Currently Absent Any Migrant Consumption Quintile Households (%) Migrant (%) (%) 1 227,732 19.0 46.5 60.0 2 234,785 12.8 7.5 19.4 3 228,143 8.9 4.9 13.3 4 213,922 8.7 4.6 13.0 5 213,366 7.3 6.3 13.1 Total 1,117,949 11.4 14.1 24.0 Note: Quintiles are calculated over yearly total household consumption net of remittances. Source: World Bank estimates using TLSS 2007 data. In both urban and rural areas among households that receive remittances, the poorest households finance large shares of their yearly consumption through remittances. Indeed, rural and urban households in the poorest quintile derive, respectively, 56 percent and 79 percent of their consumption through remittances. Across the welfare distribution, rural households are more externally dependent than urban households. Migrant households are on average more satisfied with their financial situation. Besides, they are more likely to report past financial improvements and positive expectations about the future development of their financial situation (Table 4.8). These results suggest that part of the households' improved financial situation indeed stems from the receipt of remittances. Table 4.8: Subjective Financial Indicators, Migrant Compared to Non-migrant Households Non-migrant Migrant Mean Household Household Difference Financial Satisfaction 2.44 2.50 ** Financial Situation Improved a Lot 7.7% 11.0% *** Financial Situation Expected to Improve a Lot 7.7% 10.8% *** Financial Situation Deteriorated Somewhat or a Lot 12.1% 8.4% *** Financial Situation Expected to Deteriorate Somewhat or a Lot 6.1% 4.1% ** ** Significant at 5 percent level. *** Significant at 1 percent level. Note: The financial satisfaction index ranges from 1 (unsatisfied) to 4 (satisfied). Source: World Bank estimates using TLSS 2007 data. Households receiving remittances are surprisingly less likely to spend on housing repair, but conditional on spending on this item, spent on average much larger amounts of money during the last 12 months. Controlling for other factors, only female-headed households without children substantially increase their spending on adult cloths after the receipt of remittances. For the poorest part of the population, remittances are most likely used to meet nutritional requirements, and thus do not increase spending on durables or clothing. Remittances induce substantial upward mobility among households. More than 53 percent of the poorest pre-remittance-quintile households manage to move up at least two quintiles (into quintiles three to five). Similarly, households from all other quintiles manage to improve their welfare standards. 54 D. Conclusions D1. Migration movements and profile of migrants After the breakdown of the Former Soviet Union, both internal and external migration incidence in Tajikistan increased sharply. While internal migration was mainly caused by a severe civil war between 1992 and 1997, the external migration was driven by ethnic motivations in the first years of independence but became labor-dominated soon after. The total stock of migrants was 350,000 in 2007 (January­October), of which 100,000 had already returned in 2007.53 In 2007, 5 percent of the population of Tajikistan was abroad, which equals 19 percent of the working population. About 96 percent of migrants chose Russia, and of those, more than 51 percent chose Moscow as their migration destination. Overall, 55 percent of migrants work in the construction sector. The median of migration duration of return migrants is somewhat higher for female migrants (8 months; mean 13.2 months) compared to male migrants (7 months; mean 9.5 months). Migrants on average earn about US$300 per month. Migrants are predominantly young, married males with general secondary education. Males account for 95 percent of all migrants. Around 75 percent of migrants hold a secondary education degree, while only 60 percent in the non-migrant population do (that is, non-migrants are more educated). Almost two-thirds of male migrants and three-fourths of female migrants were unemployed before leaving Tajikistan. In rural, poorer, and more remote areas, migration incidence is substantially higher than in urban areas, suggesting that migration is used as a poverty-coping strategy. In the mountainous oblast of GBAO, almost 7 percent of the population has migration experience or is abroad, while only 3.4 percent and 4.2 percent of the populations of Khatlon and Dushanbe, respectively, are abroad. Poorer households are more likely to send household members to work abroad. The largest share of migrants is employed in the construction sector (55 percent), but also in sales (9 percent) and other low-skilled occupations. Migrants in the construction sector who have returned to Tajikistan were especially exposed to illegal (that is, with no work permit) work--only 52 percent were legally employed. Illegal work was much less common among housekeepers, professionals, and agriculture workers. Average monthly earnings abroad are US$313, six times higher than the average earnings in Tajikistan (when computing the monthly earnings per working adult). Earnings differ across sector of employment, with migrants in agriculture earning only US$187 per month and professionals earning US$406 per month. Earnings abroad are positively associated with age and level of education. Furthermore, the wage premium for legally arranged jobs is around 11 percent when controlling for individual and household characteristics and for regional fixed effects. Almost all migrants remit money, either in cash (80.8 percent) or in-kind (7.3 percent). However, a limited share of migrants seems not to remit any funds, even after controlling for an initial adaptation period potentially used by migrants for the allocation of earnings. 53 Unfortunately, the TLSS 2007 provides only limited information on past migration histories and only collects information on the last migration spell. Thus, we are certainly underestimating past migration flows. 55 Upon return, migrants have lower employment rates compared to those who have no migration experience. However, return migrants are more likely to become self-employed (29.8 percent compared to 23.5 percent in the non-migrant population). Also, return migrants' net pay per month and their hourly wage rate exceed the non-migrants' monthly net pay and hourly wage rate by 35 percent and 19 percent, respectively. D2. Remittance receipt and welfare implications for households Twenty-four percent of all households have at least one migrant abroad. Rural and poorer (pre- remittance) locations are likely to have a higher share of households with migrants. At the national level, households with migrants account for 27 percent and 19 percent of households in rural and urban areas, respectively. Households with a migrant are more likely to be female-headed (30.4 percent compared to 19.4 percent in non-migrant households). Remittances have a significant bearing on household consumption. On average, urban households can afford 10 percent of their yearly consumption and rural households can afford 15 percent of their yearly consumption through remittances. One can see how deeply recipient households depend on the remittances because remittances account for 35 percent of the household total consumption. The poorest rural households finance on average as much as 80 percent of their yearly consumption through money receipts from abroad. The poorest urban households derive as much as 50 percent of their consumption through remittances. However, remittance-receiving households have only a somewhat lower poverty headcount. This can be explained by the fact that migrants are drawn from all parts of the pre-remittance welfare distribution (that is, not only poor households send migrants, although the poor are somewhat more likely to do so). Also, while some of the poorest households are highly dependent on remittances, they do not get enough in remittances to escape poverty. Migrant households are significantly more satisfied with their current financial situation than non- migrant households. Those households also report both better past welfare improvements and better expectations about future financial developments. However, remittance-receiving households are less likely to have recently invested in house repair or to have purchased cloths. However, households making these types of expenditures spent, on average, substantially larger amounts of money. Remittance receipt in Tajikistan induces substantial social mobility across the welfare distribution. More than 53 percent of households in the lowest welfare quintile before remittance receipt manage to climb at least two welfare quintiles with the help of external finance. On the downside, migrant households are extremely dependent on external finance and thus are highly vulnerable to external shocks affecting the flow of remittances. 56 Chapter 5: Social Protection in Tajikistan54 A. Introduction The purpose of this chapter is to assess the social protection (SP) system in Tajikistan and offer broad recommendations to policymakers on how to improve the system. The discussion is based on quantitative analysis of the 2003 and 2007 Tajikistan Living Standards Survey (TLSS) supplemented by discussions with key stakeholders during a visit in September 2008.55 The main focus of the chapter is on coverage, targeting, and poverty impact of social protection cash transfers and on broad options for scaling-up social assistance. The chapter considers coverage of the system followed by a discussion of targeting and the adequacy of benefits and their impact on poverty, and then discusses a recommended alternative program structure. The discussion presented here does not address broader fiscal and implementation issues of the SP system. Conclusions are briefly summarized at the end. In a context of high poverty and low domestic employment, Tajikistan runs a small SP system dominated by pay-as-you-go old-age and disability pensions with low social assistance coverage. Social protection system is dominated by social insurance in the form of pensions; the social assistance component of the SP system is small, and takes a form of a gas and electricity subsidy and a small program of allowances to poor families having children of primary school age. In addition, the United Nations World Food Program (WFP) distributes food aid. Fiscal spending on social assistance in Tajikistan is around 0.5 percent of gross domestic product (GDP), the lowest in the Eastern Europe and Central Asia (ECA) Region and less than half spent in neighboring countries such as Kyrgyzstan and Kazakhstan (Figure 5.1). Thus, social assistance spending is very low and programs are weak. 54 This chapter was prepared by Rasmus Heltberg (Senior Technical Specialist, HDV), with inputs related to the pension system provided by Ufuk Guven (Social Protection Specialist, ECSHD). 55 The author is grateful for discussions, inputs, and help from Mr. Abdurahmonov Barot, Deputy Director, Social Protection Agency; Furkat Lutfulloev, UNICEF Child Protection Unit; Emin Sanginov, Deputy Minister, Ministry of Social Protection and Labor; Vladan Milisik, Head of WFP; Nazirjanova Matluba, Head of Social Sector Department, Ministry of Finance; Pulatov Pulat Attoevich, Director, Institute of Labor and Social Protection; the European Commission; and Oleksiy Ivaschenko and Ufuk Guven. The resident mission--especially Shoira Zukhurova--arranged the visit with impeccable efficiency. 57 Figure 5.1: Public Spending on Social Assistance in Eastern Europe and Central Asia Public Spending on Social Assistance, % of GDP Croatia Bosnia-Herzegovina Hungary Ukraine OECD Uzbekistan Macedonia Russia Moldova Armenia Latvia Estonia Belarus Lithuania Bulgaria Kyrgyzstan Kosovo Serbia Romania Kazakhstan Albania Poland Georgia Turkey Azerbaijan Tajikistan - 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 Source: Nguyen, Sundaram, and Lindert (2009); and Nguyen and Sundaram (2009). The impact of the SP system on poverty and inequality is marginal, but could be improved by a new unified social assistance program offering poverty-targeted cash transfers. This is because benefits are small, coverage of social assistance is low, and targeting of old-age pensions (by far the largest program) is, as expected, marginally pro-poor.56 We propose a reform of social assistance with the goal of complementing the existing pension system with a unified social assistance benefit targeted to the poorest households. Fiscal resources permitting, the proposed unified social assistance benefit could be used to expand the SP system, achieving higher coverage of the poor. In addition, SP reform might consider providing a permanent response mechanism to a variety of shocks such as macroeconomic and financial crises, natural disasters, and climatic risks related to climate change. Concessional finance might be available to the SP system if it rises to the challenge of offering the vulnerable an effective safety net against food insecurity stemming from these aggregate shocks. B. Coverage of Social Protection In Tajikistan the main social protection (SP) programs are contributory labor pensions (which provide old-age, disability, and survivor benefits) and non-contributory social pensions. The pension system attempts to achieve both poverty alleviation and income-replacement objectives. The minimum pension is linked to the minimum wage. According to the TLSS 2007, around one-third of old- age pensioners received the minimum pension prior to its increase in August 2008; many more would receive it after the increase in minimum pensions. While the redistributive aspects mean that the system is relatively more progressive and potentially better at alleviating poverty than contributory schemes that tightly link contributions and benefits, such aspects weaken the link between contributions and benefits. 56 Although there are exceptions, individuals that contribute to the contributory scheme and eventually become eligible for a pension are not necessarily the poorest segment of the population given that they have a formal job and then a pension that they live on. 58 While coverage of old-age pensions has increased, coverage of other programs has declined since 2003. In 2007, 33.3 percent of households had one or more member receiving an old-age pension, an increase from 27.2 percent in 2003 (Table 5.1). In 2007, 5.2 percent of households received the disability pension, a decline from 8.8 percent in 2003. Various other smaller programs were received by 0.4 percent in 2007, a decline from 7.8 percent in 2003, when secondary school stipends and other programs were far more common than in 2007. Overall, the share of households covered by any SP program declined to 34.4 percent in 2007 from 38.7 percent in 2003 because of the decline of the smaller social assistance programs. Table 5.1: Coverage of Social Protection Transfers, 2003 and 200757 2007 -- Share of households receiving 2003 -- Share of households receiving Old age Disability Other All Old age Disability Other All benefits benefit benefit benefits benefits benefit benefit benefits Quintile (pre-transfer) 1 39.8 8.7 0.9 41.6 31.7 11.6 9.0 44.9 2 35.1 5.7 0.4 36.2 27.9 8.7 8.8 39.7 3 31.2 3.5 0.2 32.0 27.8 8.2 7.5 39.0 4 28.2 3.9 0.2 29.1 25.1 8.2 8.1 36.9 5 20.8 2.3 0.2 21.2 23.4 7.3 5.5 33.2 Location Urban 24.8 4.4 0.2 26.1 23.4 9.6 8.6 36.4 Rural 33.9 5.0 0.5 34.8 29.2 8.4 7.3 39.9 Poverty status (pre-transfer) Non-poor 25.0 3.2 0.2 25.8 - - - - Poor 37.2 6.5 0.6 38.4 - - - - Sex of household head Female 51.0 4.9 0.3 51.9 41.2 8.3 17.4 60.9 Male 29.2 5.3 0.5 30.4 23.8 8.9 5.4 33.3 All households 33.3 5.2 0.4 34.4 27.2 8.8 7.8 38.7 Source: World Bank estimates based on TLSS 2003 and 2007. Coverage of SP benefits and especially the old-age pension is higher among poor and low-income households, in rural areas, and among female-headed households. For example, 38.4 percent of poor households receive one or more SP benefits compared to 25.8 percent of the non-poor. And 51.9 percent of female-headed households received benefits compared to 30.4 percent of male-headed households (Table 5.1). Coverage is also significantly higher among households that have a disabled or chronically ill member, at 48.2 percent compared to 26 percent among households with no disabled or chronically ill members (Table 5.2). This reflects both the disability pension, received by 10 percent of households with a disabled or chronically ill member, and the higher incidence of illness and disability among elderly pension recipients. 57 The following types of allowances and compensations are appointed and paid against social insurance funds: (i) Pensions; (ii) Pension premium paid from the funds accumulated at individual savings accounts;(iii) Temporary disability allowance; (iv) Prenatal allowance; (vi) Family allowance (lump-sum maternity allowance, monthly parental benefit); (vii) Unemployment benefit; (viii) Funeral allowance; (ix) Spending for health improvement of workers and members of their families. 59 Table 5.2: Chronic Illness or Disability and Social Protection 2007 -- Share of households receiving Old age Disability Other All benefit benefit benefits benefits Presence of chronic illness or diability in household Yes 48.2 10.0 0.8 50.0 No 25.3 2.6 0.2 26.0 All households 33.3 5.2 0.4 34.4 Source: World Bank estimates based on TLSS 2007. The share of households with elderly members has remained constant and the increase in coverage of the old-age pension is instead caused by a sharp increase in coverage among households without elderly. In both 2003 and 2007, 27.9 percent of households contained a member of pension-eligible age, which is 63 and above for men and 58 and above for women (Table 5.3). However, while in 2003 only 4.1 percent of households without a pension-age member reported receiving the old-age pension, in 2007, this share had increased to 14.5 percent. The reasons for this are unclear--households without elderly could receive a survivors or orphans pension, or could benefit from other exemptions to the rules. Meanwhile, coverage among households containing elderly of pension-eligible age actually declined, from 87 percent to 82 percent. Again, the reasons for this decline in coverage are unclear. Table 5.3: Coverage of Old-age Pensions among the Elderly, 2003 and 2007 2003 2007 % of Household % of H ousehold households receives old- households in receives old- in category age pension ca tegory age pension Household has a member in the pension eligible age (63 and up for men, 58 and up 27.9 86.8 27.9 81.7 for women) Household does not have a pension-age 72.1 4.1 72.1 14.5 member Source: World Bank estimates based on TLSS 2003 and 2007. Old-age-pension benefits comprised the bulk (89 percent) of SP benefit receipts while the disability benefits accounted for 10 percent of the total SP benefits in 2007. Benefits from smaller social assistance programs accounted for the remaining 1 percent, down from 18 percent of the total in 2003. Other benefits comprise several very small programs many of which have shrunk between 2003 and 2007 (Table 5.4). The gas and electricity compensation (with a budget of 25 million somoni per year in 2007) reaches few of the poor. Funding is channeled to utilities to lower losses instead of reaching households directly. In other words, the population cannot make use of the amount of compensation. The compensation is paid directly to the utilities, which are supposed to provide gas or electricity for free to eligible households. Eligibility is determined according to the decisions of city and rayon commissions. However, the intended beneficiaries are often unaware that they are eligible for compensation. The Government has been taken measures to improve the mechanism of delivering the compensation payments. Apart from another program that offers allowances to poor primary school students, the country currently does not operate other social assistance programs. Some 10 million somoni are reportedly 60 channeled through school committees for the purpose of offering allowances to poor primary school students. No information is available on the effectiveness of this program and the program is hard to assess via the TLSS 2007 survey. Table 5.4: Distribution of Total Social Protection Spending across Programs, 2003 and 2007 O ld a g e D is a b ility O th e r b e n e f it b e n e f it b e n e fi t s 20 0 7 89% 10% 1% 20 0 3 63% 20% 18 % N o te : O th e r b e n e f its in 2 0 0 3 in c lu d e s a l a rg e r n u m b e r o f p r o g r a m s t h a n i n 2 0 0 7 , f o r e x a m p l e s e c o n d a ry s c h o o l s t i p e n d s a n d s u r v i v o r p e n s i o n s . T h e s e a r e g ro u p e d u n d e r o th e r . Source: World Bank estimates based on TLSS 2003 and 2007. C. Incidence of Benefits The incidence of social protection (SP) benefits is mildly pro-poor, reflecting a strong redistribution through pensions and disability payments. Results on incidence of SP benefits across population groups for all SP programs combined are similar to those for the old-age pensions alone, reflecting that old-age pensions dominate the SP system as the largest program by far. About 28 percent of old-age pensions go to the bottom 20 percent of households and 49.4 percent go to the bottom 40 percent of households (Table 5.5). This is expected despite the fact that the re-distributional aspects of the Tajik pension scheme are contributory and benefits are linked to wages, although the link is weak. Disability pension benefits are highly pro-poor in their targeting, with 40.6 percent going to the first quintile and 57.4 percent going to the 1st and 2nd quintiles combined. The targeting of all SP benefits improved during 2003­07, while coverage declined. Targeting performance of disability pensions improved the most, from nearly 27 percent of benefits going to the 1st quintile in 2003 to more than 40 percent in 2007. The improvement in the incidence of the disability pension occurred at the same time that its coverage shrank. Table 5.5: Benefit Incidence and Targeting, 2003 and 2007 Share of benefits going to households in... 2007 2003 Old age Disability Other All Old age Disability Other All benefits benefit benefit benefits benefits benefit benefit benefits Quintile (pre-transfer) 1 27.9 40.6 38.7 29.2 24.7 26.9 26.1 25.4 2 21.5 16.8 23.2 21.1 18.0 19.1 13.6 17.4 3 21.0 17.7 2.2 20.6 20.7 18.1 20.5 20.1 4 16.7 15.8 12.8 16.6 17.7 21.6 21.8 19.2 5 12.9 9.2 23.0 12.6 18.9 14.3 18.0 17.9 Poverty status (pre-transfer) Non-poor 39.0 32.9 36.7 38.4 - - - - Poor 61.0 67.1 63.3 61.6 - - - - Sum 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Source: The World Bank estimates based on TLSS 2003 and 2007. 61 The targeting effectiveness of Tajikistan's social assistance (SA) programs does not compare favorably with that of other countries. Internationally, good programs deliver at least 40 percent of their SA benefits to households in the bottom 20 percent. Tajikistan delivers only 23 percent of their SA benefits to households in the bottom 20 percent (Figure 5.2). Figure 5.2: Comparative Targeting of Social Assistance across Countries Targeting Accuracy of Social Assistance: % of Total Benefits Received by Poorest Quintile (All Non-Contributory Transfers) 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Source: The World Bank (ECSHD) database on social assistance. D. Adequacy and Poverty Impact of Social Protection (SP) Benefits SP benefits are modest, with average total benefits per household, conditional on receiving one or more benefits, of just 38.6 somoni per month, or 5.5 percent of average total household expenditures. This average includes all benefits received by the household. It corresponds to around US$ 9 per month per household at market exchange rates (Table 5.6). Average old-age and disability pensions were 36 somoni and 26 somoni per month per household, respectively. These data are from before the pension increase in August 2008 (more on this below). Averaging these benefits across all households, both receiving and not, gives 12 somoni per household per month. SP benefits account for just 1.2 percent of total household expenditures in Tajikistan. The 2007 survey indicated that SP benefits declined in recent years relative to beneficiary households' total expenditures. Old-age-pension benefits constituted just 5.1 percent of beneficiaries' average expenditures in 2007, although somewhat more (8.3 percent) for beneficiaries in the poorest quintile. Other benefits provided even lower shares of household expenditures (3.6 percent for disability pensions on average). Benefit adequacy has declined in recent years. In 2003, old-age pension benefits constituted 6.6 percent of average expenditures and 11.8 percent for those in the poorest quintile (Table 5.7). The decline in benefit adequacy is even larger for disability pensions; this decline is disappointing given the pro-poor targeting performance of the disability pension. 62 Table 5.6: Average Size of Social Protection Benefits and Expenditures, 2007 (Somoni per household per month) Old age Disability Other All Per capita Total benefit benefit benefits benefits expenditures household expenditures (Beneficiaries) (All households) Quintile (pre-transfer) 1 38.8 29.0 16.0 43.6 78 558 2 34.4 19.4 18.7 36.6 112 771 3 36.5 27.0 4.0 38.9 143 969 4 32.8 24.7 20.0 35.3 188 1108 5 33.9 27.8 34.9 36.2 371 1582 All 35.6 25.7 18.1 38.6 178 997 Source: World Bank estimates based on TLSS 2007. Benefits had a low impact on poverty and well-being. A rough benchmark of adequacy that is sometimes used is that in order to be meaningful, benefits should correspond to at least 20 percent of beneficiaries' total household expenditures. A significant increase in minimum nominal pensions (made effective in August 2008) has improved the adequacy of pension benefits for recipients (more on this below). Table 5.7: Adequacy of Social Protection Benefits (benefits as percent of household expenditures conditional on receiving, 2007 and 2003) 2007 2003 Old age Disability Other All Old age Disability Other All benefit benefit benefits benefits benefit benefit benefits benefits Quintile (pre-transfer) 1 8.3 5.5 1.8 9.1 11.8 10.9 11.8 13.5 2 4.6 2.2 1.8 4.8 5.7 5.6 4.0 6.2 3 4.0 3.0 0.4 4.3 5.4 4.6 4.8 5.7 4 3.5 2.5 1.3 3.8 4.9 4.4 4.3 5.2 5 3.5 2.0 1.3 3.6 4.2 2.5 3.5 4.1 Poverty status (pre-transfer) Non-poor 3.5 2.4 1.2 3.7 - - - - Poor 6.2 4.2 1.7 6.7 - - - - All beneficiary households 5.1 3.6 1.5 5.5 6.6 6.1 6.0 7.2 Source: World Bank estimates based on TLSS 2003 and 2007. There are no major issues with SP benefits arrears. The TLSS 2007 contains detailed questions on arrears in receiving benefits. According to household responses, arrears are relatively few and, when they occur, last just a few months. This indicates that the system has preserved a basic implementation capacity to deliver the benefits. The Ministry of Labor and Social Protection (MLSP) has made several changes to the pension system in recent years, including a substantial raise in nominal pension benefits. The Social Protection Fund was converted into the Social Protection Agency under the MLSP in 2006. Minimal pensions (both social and contributory) were raised from 20 somoni to 60 somoni per month effective August 2008. The increase in minimum pensions was linked to an increase in the minimum wage of the same amount. Pensions above the minimum were also adjusted with the size of the raise depending on various factors. Many pensioners are reported to have received an increase of 24 somoni per month. 63 Calculations indicate that recent adjustment in pension sizes has doubled average pension benefits from 38 somoni to 75 somoni per month per receiving household. This adjustment has gone some way toward addressing the inadequacy of the benefits and the very low impact on the poverty of the beneficiaries (see below). A simulation indicated that the recent pension increase has doubled the ratio of benefits to household expenditures from 5 percent to 10 percent on average. The same simulation also suggested that the benefit adjustment has not significantly influenced the targeting performance of the SP system. While the widespread migration offers many opportunities for income and employment, it also provides a challenge to the Government in that migrants do not contribute to the SP system. According to the TLSS 2007, Tajikistan has 1.8 million domestically employed workers and at least 350,000 labor migrants are outside the country.58 The economically active population as a share of the total working-age population is only 43 percent, which is low by international comparison. A low activity rate combined with migration means that relatively few domestic workers contribute to the social security system or pay taxes to help finance social assistance. The migrants do not contribute to the system but expect a pension when they return home to retire in Tajikistan. This is an understandable source of concern for the Government but not one that is easily addressed. Some countries that have tried voluntary pension savings plans for informal and migrant workers have found that uptake is often rather limited. The fact that migrant workers expect to receive a pension when they retire places fiscal strains on what is, at least in part, a contributory system, and calls for the reform of the current system. Aiming, among other things, to bring the migrant workers into the pension system, the MLSP has drafted a law on Social Insurance and Pensions proposing to introduce a Defined Contribution system, apparently inspired by the Russian system. Based on accumulation of private savings subject to financial returns, this would mark a major reform of the current pay-as-you-go system. As other countries have experienced, the introduction of such system needs to go hand in hand with significant institutional and administrative capacity necessary for providing an efficient framework for private pension fund operators. Given the small size of SP benefits, it is not surprising that the SP system has negligible impacts on poverty and inequality. The estimations suggest that SP transfers reduce poverty headcount from 54.8 percent to 53.1 percent, a reduction of 3 percent, or 1.7 percentage points. The estimated reduction in the poverty gap is 5.1 percent (Table 5.8). The estimated impact of transfers on inequality is also very low, suggesting a reduction in the country's already low Gini Coefficient of inequality of just 1.5 percent, or 0.4 percentage points--from 29.4 percent to 29.0 percent. Simulations of the August 2008 increase in minimum pensions suggests a modest improvement in the poverty impact of SP benefits with the impact of transfers on poverty headcount increasing to 5.3 percent, or a 2.8 percentage point reduction in poverty.59 58 Migration is discussed in detail in Chapter 4. 59 These estimations assume that all transfers are consumed and that transfers have no offsetting value on labor supply and other income sources. The estimated impacts of transfers would be lower than these estimates to the extent that households save a portion of transfers or reduce their labor supply. 64 Table 5.8: Simulated Impact of Social Protection Benefits on Poverty 2007 Poverty headcount Poverty gap Before transfers After transfers Before transfers After transfers Urban 50.0 49.3 16.2 15.4 Rural 56.4 54.4 15.7 14.9 Total 54.8 53.1 15.8 15.0 Reduction 3.0% 5.1% Assuming transfers are spent entire on consumption and do not influence labor supply Source: Staff estimates based on TLSS 2003 and 2007. E. Toward a More Comprehensive and Effective Social Protection (SP) System The weaknesses of Tajikistan's SP system--and especially the absence of efficient social assistance--have been vividly illustrated in the current financial crisis where the country (and its donors) finds itself lacking an effective SP mechanism to protect the poorest. Whereas many other countries have been able to respond to food and financial crises by improving and scaling-up preexisting safety net programs to cover more poor and vulnerable households, sometimes using foreign aid, Tajikistan was in a more difficult position since no such program exists in Tajikistan. The food aid distributed by the WFP has instead been a major crisis response. The international experience is that good programs take 12 to 18 months to set up and that there is a large payoff to having them in place before crises. Once set up, programs can be helpful in response to a range of shocks such as macroeconomic, natural disaster, and climatic shocks, as long as programs have a flexible delivery mechanism that is able to increase the number of beneficiaries in response to needs. Good SP programs might also help Tajikistan adapt to ongoing climate changes and use concessional climate change funding as a potential source of finance for SP. Tajikistan was recently invited to participate in a new Pilot Program for Climate Resilience (PPCR), which will provide significant funding for scaled-up action in integrating climate resilience in national development planning. The PPCR is part of the new multibillion Climate Investment Funds housed at the World Bank. While there is no blueprint for how countries will use PPCR resources, the aim is to give the poorest and most vulnerable countries an opportunity to transform development planning by integrating climate considerations and to promote innovative approaches to address climate resilience in development planning and investments. Key themes expected to gain support under the PPCR include community- based risk reduction measures and increasing food security. A useful starting point for improving Tajikistan's SP system is to consider a single, unified social assistance program of some scale--rather than a myriad of smaller, overlapping programs--with competent staff trained in targeting, delivery, and handling of complaints. Expanding social assistance does entail short-term fiscal costs, but this would only help bring social assistance expenditures more in line with regional averages. It is possible that foreign donors would contribute to financing the proposed expansion. The key up-front investment is in administrative capacity to target and deliver the benefits. Experience shows that this is much more easily done for a single, unified program than for a myriad of minor benefits, and for the long term rather than for ad hoc crisis response interventions.60 60 Mexico's Progresa (now Oportunidades) and Indonesia's Kecamatan Development Program (KDP) are examples of highly successful poverty programs that were set up in times of crisis and subsequently expanded to become important permanent elements of social policy. 65 Objectives and target group are some of major issues that Tajikistan will confront if it chooses to set up a unified social assistance program. For example, the objective of a social assistance cash transfer program may focus on alleviating general poverty and food insecurity, or more on child poverty and human development. If general poverty is the key concern, unconditional cash transfers targeted to the poorest may be a good choice. If child poverty and human development is the paramount concern, international evidence is clear that conditional cash transfers (CCTs) can be a highly effective tool. CCTs in many countries give cash transfers to parents--usually mothers--as long as their children attend school and/or visit health centers for checkups and vaccinations. CCTs induce behavioral change and help build human capital among the poor, which can fight intergenerational poverty. Targeting method is another major issue when setting up a new social assistance program. No safety net can ever be perfectly targeted. There will always be targeting errors, large or small, which program design should take into account and seek to minimize. The literature refers to two different kinds of targeting problems: inclusion errors--some non-poor are included in the program, and exclusion errors--some poor are left out of the program. In addition to inclusion and exclusion errors, there are also other factors that reduce the share of program resources being transferred to the poor, which program design also needs to take into account, namely corruption/leakage--resources not reaching any beneficiary--and administrative costs. Safety net schemes worldwide use several different methods of targeting cash benefits, and there are a lot of good models that Tajikistan can follow. Means testing involves testing an individual's income and assets, but is not relevant for a largely informal economy where income is impossible to verify, as in Tajikistan. In this situation, many countries, for example, Armenia, Georgia, Russia, and many countries in Latin America, use proxy means testing based on observable characteristics correlated with poverty to identify the poor (for example, area of residence, housing characteristics, durable goods, and so forth) (Box 5.1). This mode of targeting may be a good possibility for Tajikistan. Categorical targeting is a simpler targeting method that uses a single characteristic (for example, old age, disability, or number of children), but is unlikely to be relevant since Tajikistan already has benefits going to many disabled and elderly. Community-based targeting has been found to perform very well in analysis of global targeting experience, particularly in countries where communities have a long tradition of collaborating in various spheres of life (for example, Albania and Uzbekistan). Community-based targeting may also be relevant to Tajikistan. Finally, geographic targeting is often used either in isolation or in combination with one of the other targeting methods, whereby residents in poorer areas are eligible for assistance and those in better-off areas are not. Simulations suggest that in Tajikistan geographic targeting could be a very promising option. Simulations of geographic targeting and proxy means testing and combinations of the two indicate that geographic targeting in isolation would most precisely identify the poor and have the largest impact on poverty (Annex 12). An additional benefit of geographic targeting is its administrative simplicity. The WFP practices a variant of geographic targeting in its food distribution programs in the country. The four main donors active in the SP sector have adopted a functional division of responsibilities along the lines of comparative advantages. The World Bank leads on pensions and has financed a functional review of SP in the country. The European Community (EC) focuses on social assistance and has approved technical assistance and a 5-million-euro budget to support an operation focused on the SP system. This budget support is conditional on progress on implementing a policy matrix for SP that has been signed by the Government. The policy matrix covers social assistance, social care services, a labor market study and strategy, poverty statistics, and public sector financial management of SP. It does not cover pensions. The United Nations Children's Fund's (UNICEF's) Child Protection Unit focuses on disability and social care and (with EC funding) helps the Government set standards for social care and training social workers. The WFP distributes food aid in food-deficit areas. 66 Box 5.1: Proxy Means Testing One tool for better targeting that may be considered is a regression-based proxy means test (PMT). This targeting tool is being used by a growing number of countries, including Chile, Colombia, Georgia, and Mexico, with promising results. Proxy means tests rely on a composite formula based on observable household characteristics (assets, demographics, education, occupation, and housing), to estimate household welfare, inferring from the statistical relationship between observed characteristics and the observed poverty status of households. The weights are usually estimated using regression techniques based on the observed characteristics and the observed poverty status of households in a household survey. Households whose estimated income falls below a predetermined threshold are eligible for transfers. To preempt households from misreporting their characteristics, proxy means tests need to be based on characteristics that can be verified fairly easily. The Tajikistan Living Standard Survey (TLSS) 2007 data would be well suited to develop and test a proxy means formula taking the following steps: 1. A regression is run in which the welfare measure, per capita expenditures, is regressed on a number of household characteristics (demographics, education, and so forth). 2. The estimated parameters of the expenditure regression are used to predict the welfare level of all households--the estimated regression parameters comprise the proxy-means-testing formula. 3. The predicted welfare level is compared to a cutoff point to determine if the household is eligible. 4. The targeting efficiency of the proxy-means-testing formula is evaluated--using the same survey data-- by comparing households deemed eligible for assistance according to the formula and households with actual expenditures below the cutoff. A targeting success is when a household is placed on the same side of the cutoff by both the formula and the survey data. Targeting errors occur when formula and observed data disagree on which side of the cutoff to place a household. Households that are predicted to be non- poor according to the formula but whose actual survey welfare is below the cutoff are considered to be undercovered. Households predicted eligible for assistance by the formula but actually living above the cutoff are said to represent leakage. A recent study on Latin American and Caribbean countries found that PMT performed well in terms of targeting outcomes, cost-efficiency, and transparency. PMT systems in Latin America have generated targeting incidence outcomes that approximate the impressive record of verified means tests for a mere fraction of the cost. Between 60 and 90 percent of the benefits of proxy-means-tested programs in Chile and Mexico are received by the poorest 40 percent of households in those countries. Moreover, the costs of these systems are relatively low (ranging from US$2.3 to US$8.4 per interview in Latin America, or 9 percent to 34 percent of comparable interview costs for verified means tests in the United States), and its administrative requirements are relatively manageable. Finally, the PMT systems in several Latin American countries also rank fairly high for transparency. F. Conclusions The main finding of the analysis is that Tajikistan operates a weak social protection (SP) system with old-age and disability pensions as its cornerstones and virtually no social assistance. Benefits are small and impacts on poverty are extremely modest. Fiscal spending on social assistance is the lowest in the Eastern ECA Region. Illustrating the weakness of Tajikistan's SP system, especially social assistance, is the lack of a mechanism to protect the poorest during the current financial crisis. Whereas many other countries have responded to the crisis by improving and scaling-up safety net programs, no such programs exist in Tajikistan. Therefore, the country has been unable to leverage SP for crisis response. A well-designed, scaled-up SP system could help the country respond to a variety of shocks and crises, including adverse climatic events linked to climate change. 67 The chapter proposes to begin the scaling-up of the SP system by introducing a new unified social assistance program offering poverty-targeted cash transfers. The preference is for a single, unified program of some scale rather than a myriad of smaller overlapping programs. Under this proposal, the electricity and gas compensation would be discontinued and its budget rolled into the new program. Over time, investments could be made in staff training, targeting, delivery, and handling of complaints. As discussed, objectives, target group, and targeting method are major choices confronting the country if it chooses to set up a unified social assistance program. If general poverty is the key concern, unconditional cash transfers targeted to the poorest may be a good choice. If building human capital of poor children is the main concern, then conditional cash transfers could be an effective tool, particularly if linked to secondary schooling. While a range of different methods for targeting cash benefits are available, the most promising for Tajikistan appears to be geographic targeting. The World Bank and other agencies have technical capacity to design the program of targeted social assistance based on agreed criteria. While the proposed scaling-up of social assistance would entail some fiscal costs,61 these costs need to be considered in light of the averted poverty and human capital losses. SP expenditures should be seen as an investment in a long-term effort to fight poverty and offer citizens protection against a range of shocks and crises. In the future, such a program will offer a useful way to protect food security in times of shocks, be it shocks from a macroeconomic crisis, natural disasters, conflict, or adverse climatic events stemming from global climate changes. 61 The calculations indicate that the social assistance program that would cover extreme poor (about 1 million people) and pay each household the benefit value equal to about 10% of the extreme poverty line (about 10 Somoni) per month, would cost the budget about 1% of the GDP. 68 Chapter 6: Education in Tajikistan ­ Progress but Many More Challenges Ahead62 A. Introduction The tumult after independence and subsequent civil war had a significant negative impact on the Tajik education system. Funds for education diminished significantly and years of civil war left infrastructure unmaintained for nearly a decade, while teachers migrated abroad, leaving classrooms short of teachers. Since the end of the civil war in 1997, the Government of Tajikistan (GOT) has been reestablishing the education system, including improving the infrastructure, and, recently, starting reforms to improve the financing system and the relevance of curriculums. This chapter will assess: (a) how education is financed from public and private funds, (b) the trends in enrollment and exclusion, and (c) what recent indicators of quality in the education system can tell us. The chapter makes particular use of the 2007 Tajikistan Living Standards Survey, which provides key insights on education, and ends with observations and recommendations for further improvements of the education system. Tajikistan's education system consists of four types of schools: (a) preprimary schools, (b) general schools, (c) specialized secondary and professional training schools (including vocational schools), and (d) higher education institutions (see Table 6.1 for more details). General schools include primary, lower secondary, and upper secondary grades. The specialized secondary schools also offer secondary and upper secondary education, but generally offer vocational and technical type of education. Preprimary education, which can begin as early as three months of age, is only partially intended to prepare children for entry into the more formal education system, and functions more like daycare. All education system supervision falls under the Ministry of Education (MOE), and provision is overwhelmingly done by public institutions. The responsibility for overall educational policy and curriculums for all schools fall under the MOE.63 The MOE, in consultation with relevant oblast and rayon authorities, ensures that the minimum standards are maintained. Since 1994, licensed, fee-based private schools have also been allowed. These private schools also fall under the responsibility of the MOE, but education so far is overwhelmingly provided by public institutions. In 2007, 89 percent of preprimary children were enrolled in public institutions, while the corresponding shares for general education, specialized secondary, and higher education were 99, 97, and 98 percent, respectively (TLSS 2007). 62 This chapter was prepared by Thomas Pave Sohnesen (Consultant, ECSHD). 63 Including schools that previously fell under other ministries, such as vocational schools, which were under the Ministry of Labor. 69 Table 6.1: Tajik Education System Duration of Level of Education Age Educational Institutions Studies 1­3 years, Preprimary 1­6 (7) Kindergarten/nursery 3­6 years General education: 11 years 7­18 · Primary 4 years 7­11 General education schools, · General basic 5 years 11­16 gymnasiums, lyceums · General secondary 2 years 16­18 Professional education: Vocational schools, centers, technical 1­4 years From age 16 · Primary colleges, colleges, special secondary 2­4 years From age 16 · Secondary schools, universities, academies, 4­6 years * · Higher institutes Master courses, postgraduate courses, Post-diploma education doctorate courses Additional (extra) education ** * Conducted on the basis of general secondary, primary, and secondary professional education. ** Conducted in regular schools of general and professional education, outside main educational curriculums, or in the establishments for additional education (small research academies, palaces, stations, clubs, centers, art and music schools, and so forth). Source: MOE 2005. B. Public and Private Expenditure on Education Households finance roughly half of all educational expenditures in general and specialized secondary education, and the majority of education expenditures in higher education. Private households, for all levels of education, spend at least as much per student as the Government does on tuition, uniforms, repair of buildings, textbooks, stationery, and meals and lodging. At the level of general education and secondary specialized education, the Government and private households spend about equal amounts per student, that is, around 145 somoni (US$ 33) per student in general education and about 345 somoni (US$ 80) per secondary specialized student a year. Once students progress beyond secondary education, households have to cover the lion's share of expenditures. On average, higher education students pay 86 percent of educational expenditures (Annex 13, Table A13.1). That private expenditure makes up roughly half of the total expenditure on general education is not new. The same was found in 1999 (World Bank 2008b). The issues of public and private financing are discussed in a greater detail below. B.1 Public expenditure on education Public funding has increased for all levels of education in recent years, but it is still estimated to be insufficient for reaching the GOT's goals for the education sector. Since the implementation of the Poverty Reduction Strategy Paper (PRSP) in 2003, public spending on education has increased steadily to 22 percent of total government expenditure in 2007, up from 12 percent in 2003. The increase is equivalent to a tripling of expenditures per capita, from US$5 in 2002 to US$18 in 2007. However, despite the Government's commitment and increasing budget allocations, the public expenditures currently make up only 3.4 percent of gross domestic product (GDP). Estimates have shown that in order to meet the minimal needs for maintenance and development of the education system, the average annual budget share for education should reach 6 percent of GDP (World Bank 2008b). Accordingly, the GOT's official goal, set forth in the PRSP, is 6 percent of GDP. 70 Public financing for education comes largely from local governments (rayons [districts] and jamoats [municipalities]). General education is mostly regionally financed, while specialized secondary and higher education is mostly federally funded. In 2007, 81 percent of total public educational funds came from local government compared to the 19 percent provided by the federal government. For preprimary and general schools, local governments provided 100 percent and 97 percent of the financing, respectively, while the federal government financed most of specialized secondary and higher education (World Bank 2008b). in the current institutional setup leads to unequal financing of general education and a lack of oversight. Since local governments finance most expenditure on general schools and have significant discretionary power in allocating resources, resources allocated to each school vary significantly. The average per capita spending was found to be 7.7 times higher in the rayons with the highest expenditure per capita compared to the lowest (World Bank 2008b). The problem is exacerbated because local governments are not expected by law or practice to keep track of how funds are allocated among schools, and generally do not report to the federal government how much funding they provide for what. With this setup, the federal government has limited knowledge and oversight of total resources flowing to each school and therefore cannot supervise, regulate, or ensure adequate funding for all schools. The state does provide extra funding for schools with perceived extra needs, but since there is limited oversight there is no guarantee that these funds are well targeted. Consequently, there is only a limited link between national educational policies and finance for schools. Resource use could be significantly improved by linking planning and financing to incentives. Most local governments and schools have no medium-term budgeting system in place and schools are allocated a budget on a year-by-year basis based on past needs. The current financing system leaves schools with limited autonomy and provides little incentives for schools to shift resources around to maximize outputs. That there is room for efficiency improvement is indicated by the lack of correlation found in the World Bank (2008b) study between funding received and school output. Current MOE reforms are improving resource management, including implementation of per capita financing. A recent World Bank public expenditure review (World Bank 2008b) analyzed the many challenges to achieving a transparent, policy-based budgeting system. It is noteworthy that the MOE is the first line ministry to undertake reforms, and serves as a pilot ministry for improvement for other ministries. Since 2005, a pilot project has addressed part of these challenges by gradually implementing per capita funding for schools. In addition to the per capita financing, the pilot also gives schools autonomy over their spending. The funds are transferred directly to the schools as a lump sum and the schools themselves decide how to spend the funds on salaries, utilities, maintenance and repairs, textbooks, and other expenses. This approach is a first important step in the right direction to ensuring better allocation of funds. Very little public expenditure on higher education reaches the poor. Public higher education institutions are free of charge for about half of the students and are paid for through tuition fees by the other half. Due to high official or unofficial costs, almost all higher education students come from a wealthy background. Seventy-two percent of students come from the two first quintiles, while only 5 percent of students come from the first quintile (Figure 6.1). The few students with a poor background that were enrolled in higher education in 2006/07 were slightly less likely to pay tuition fees compared to students from a richer background. 71 Figure 6.1: Distribution of Enrolled Students across Income Quintiles, 2007 100 18 20 23 32 29 Richest 80 50 19 20 31 Rich 60 22 23 21 21 Middle 40 8 21 16 22 19 22 20 Poor 20 16 19 15 20 21 19 Poorest 12 12 8 0 5 Preprimary Primary Basic Secondary Education Specialized Secondary General Higher Source: TLSS 2007 and authors' calculations. While mildly progressive, scholarships could reach more of the poor through better targeting. Less than 5 percent of students across all education levels receive scholarships. Scholarships are most common for students at secondary specialized schools (16 percent) and higher education institutions (15 percent) compared to general education (4 percent). Since most of the students at specialized secondary and higher education schools come from better off families, they also received most of the scholarships. A similar pattern is seen at secondary specialized schools. The scholarships in general education seem better targeted at the poor even though they are far from perfect. The poorest quintile received 29 percent of the scholarships compared to 14 percent for the richest quintile. Unfortunately, due to data limitations, we are not able to tell which type of scholarship students received. Hence, we do not know who gave the scholarship (the state, the local government, the school, other private funds) and with what justification the scholarship was given. For instance, federal-state scholarships for higher education go to both students with excellent performance and vulnerable students such as orphans. Considering the overall nonprogressive profile of scholarships, a retargeting of more scholarships to needy students could be considered a way to reduce poverty and improve equality of enrollment across wealth groups. B.2 Private expenditure on education Household expenditures on education as a percentage of total consumption almost doubled between 1999 and 2007, and urban and rich spend more than rural and poor. Private expenditure on education as a share of overall household expenditure seems to have almost doubled from 2.4 percent in 1999 to 4.3 percent in 2007. Education seems to have a higher priority among rich and urban households than poor and rural households since they spent more on education. Urban households spent on average 15 somoni (US$4.4) a month on education, equivalent to 5.2 percent of total expenditures, which is significantly higher than rural households, which spend 8 somoni (US$2.3) a month, equivalent to 3.9 percent of total expenditure. The higher expenditure of urban households as a share of total expenditure and in absolute value is true for almost all consumption quintiles. In addition, rich households spend more than poorer households both as a share of total expenditure and in absolute value (TLSS 2007). 72 General education Although public education is tuition-free in Tajikistan, the survey reveals that some public general education institutions are charging tuition fees, particularly in Dushanbe and other urban areas. According to the Tajik Constitution, basic education (grades 1­9) is compulsory and free of charge for all children. In addition, the state guarantees free general upper secondary education (grades 10­11) and secondary specialized and higher education on a competitive basis in public education institutions. However, some basic education and general secondary education schools do charge tuition fees (Figure 6.2). On average, 13 percent of general education students pay tuition fees. There are wide geographic disparities in this practice since tuition fees are mostly charged in Dushanbe and other urban areas. Forty- nine percent of students in Dushanbe reported that they paid general education school fees or tuition, while in other regions this share was under 15 percent (Figure 6.2). Figure 6.2: Percentage of Students Paying Tuition Fees 80 80 72 70 70 54 60 49 51 49 52 60 49 50 44 50 38 39 50 38 36 40 40 26 30 24 23 23 30 18 13 17 12 13 15 20 12 12 9 12 20 8 6 4 10 10 0 0 General education Secondary Specialized Higher Education General education Secondary Specialized Higher Education Education Education Poorest Poor Middle Rich Richest Dushanbe Sogd KHatlon RRP GBAO Source: TLSS 2007 and authors' calculations. Basics such as uniforms and stationery supplies make up the bulk of private expenditures in general education. Despite expenditures on tuition fees, textbooks, and infrastructure, most private expenditures go to basics such as uniforms and stationery supplies. On average, 47 percent of student expenditures on education are for uniforms and 8 percent are for stationery supplies, which translates, respectively, into 90 somoni (US$26) and 16 somoni (US$4.6) (Annex 13, Table A13.2). In addition to tuition, private households also directly contribute to educational infrastructure and textbooks. Among students in general education, 85 percent had expenditure on textbooks and other instruction material and 72 percent had expenditure on repair of buildings and similar expenditures. Hence private households are compensating for things that the state has committed itself to provide but has not delivered. On average, each student paid 16 somoni for textbooks and 8 somoni for building repairs (Annex 13, Table A13.2). B.3 Specialized secondary and higher education Twice as much in private expenditure is spent on specialized secondary education as on general education, and private expenditure on higher education is six times as much as what is spent on general education. On average, students in general education spend 146 somoni a year compared to 345 somoni and 836 somoni on specialized secondary and higher education (US$42, US$100, and US$243), 73 respectively. Like for general education, school uniforms are still a significant expenditure at specialized secondary and higher education. However, meals, lodging, and tuition fees at this level are the largest expenditures. Tuition fees are very common, with significant variation in use across oblasts. Poorer students are less likely to pay tuition, but targeting is far from perfect. Twenty-three percent of students in specialized secondary education and 50 percent of students in higher education pay tuition fees. However, the likelihood of paying tuition fees depends on which oblast you study in. For instance, 72 percent of students in higher education in Sogd pay tuition fees compared to 38 percent in Dushanbe. Hence, the picture is reversed from general education, where students in Dushanbe were most likely to pay tuition fees. Poor students are less likely in both specialized secondary and higher education to pay tuition fees than richer students. Among students from the poorest quintile, 9 percent and 39 percent paid tuition fees at specialized secondary and higher education levels, respectively, compared to 38 percent and 52 percent, respectively, among the richest quintile. However, almost all students in higher education come from the richer quintiles, as we will see in the next section. C. Enrollment, with a Poverty and Gender Perspective The disruption of the education system following independence and the civil war resulted in younger generations being less educated than older generations. A visibly larger share of younger generations has obtained only lower levels of education (primary or lower secondary) and a significantly smaller share has obtained specialized secondary education (Figure 6.3). Enrollment declined during the civil war, but some of this dent in the stock of education could also be attributed to migration. There was significant permanent out-migration during those years, and anecdotal evidence indicates that migration was more prominent among the more highly educated. A larger share of older generations (over age 50) also seem to have higher education, which likely is due to the emphasis on education during the Soviet era. Figure 6.3: Educational Attainment Distribution by Age (percent) 100 1 2 7 10 11 12 12 14 11 16 17 16 90 7 10 13 80 15 18 21 12 17 19 20 70 54 60 59 32 50 58 54 40 32 40 61 59 57 53 30 24 36 20 22 22 20 19 10 19 10 12 16 9 10 7 6 4 4 4 0 2 1 1 2 2 16-20 21-25 26-30 31-35 36-40 41-45 46-50 51-55 56-60 61-65 all none Primary Basic Secondary General Secondary Specialized Higher Education Notes: Highest degree obtained for those currently not studying. Source: TLSS 2007 and authors' calculations. Even with an increasing student body, Tajikistan managed to sustain high enrollment rates at all education levels from 2000 to 2006. The education sector has been greatly challenged by demographic transition. Between 1990 and 2000, it is estimated that the population aged 5­19 increased by 27.5 percent. From 2000 to 2006, the same population increased by another 5.4 percent. This obviously put a lot of pressure on the education system, which had to cope with a large number of additional students. 74 Nevertheless, even with the increase in the student body from 2000 to 2006, enrollment levels generally increased between 2000 and 2006. Gross preprimary enrollment went from 7.3 percent to 9.1 percent, net primary enrollment increased marginally from 96 percent to 97 percent, net secondary enrollment rose from 71 percent to 80 percent, and gross tertiary enrollment went from 14 percent to 18 percent (World Development Indicators). Fortunately, the increasing demographic pressure on the education system has halted and is expected to decrease from 2030 onwards. From 2005 until 2030, the population aged 5­19 is largely expected to be steady, while after 2030 the population is expected to decline again (Figure 6.4). Figure 6.4: Population Aged 5­19, Thousands Change in Population Year Total Population (percent) 1990­2000 27.5 3000 1990­2005 34.4 2000 1000 2000­2005 5.4 0 2005­2030 4.1 90 00 10 20 30 40 50 19 20 20 20 20 20 20 2030­2050 -17.4 Source: UN population data and authors' calculations. Despite progress, Tajikistan's enrollment rates at certain education levels are lower than other countries in the ECA Region, but are also ahead of other countries at the same GDP per capita level. In 2006, Tajikistan ranked last in preprimary enrollment, second to last in secondary enrollment, and third to last in tertiary enrollment among ECA countries (World Development Indicators). It should, of course, be kept in mind when comparing Tajikistan to the region that Tajikistan is the poorest country in the region, with a per capita GDP Purchasing Power Parity of US$1,609 in 2006. This is only 15 percent of the average GDP per capita in the ECA Region, or put differently, the average GDP per capita in the ECA Region is almost seven times higher then GDP per capita in Tajikistan. This level of GDP per capita is around the average for Sub-Saharan Africa. Hence, conditional on the level of development, Tajikistan's enrollment levels in primary, secondary, and tertiary education look favorable. C.1 Preprimary enrollment Preprimary education in Tajikistan encompasses both nurseries/kindergartens for children aged 1­3 and kindergartens for children aged 3­6. The latter group is more relevant from an educational point of view, so this section will focus on them.64 Low attendance in preprimary school is likely due to limited supply, particularly in rural areas. Enrollment in preprimary education is relatively low compared to ECA countries or even to other countries with similar GDP per capita levels (World Development Indicators). This is particularly true in rural areas, where less than 3 percent attend preprimary education. It is more common in urban areas where almost one in five (18 percent) attend. Predictably, preprimary enrollment is most common among the richest 20 percent of the population, but children from all consumption quintiles enroll. The difference between urban and rural enrollment can be partly explained by access to preprimary schools. Forty-seven 64 The Tajikistan Living Standard Survey (TLSS) 2007 has data only on preprimary education for children aged 3­5. 75 percent of the urban population lives in an area with a public preprimary school (kindergarten), while only 27 percent of the rural population lives in an area with a public preprimary school. However, access can only partially explain low preprimary enrollment, since many do not use preprimary facilities even when they are available. In locations that had a public kindergarten, only 7 percent of rural and 26 percent of urban households enrolled their children.65 There are urban-rural differences in the reported reasons for non-enrollment. Sixty-three percent of rural households with preprimary-school-age children state that the children were not enrolled because there were no preprimary schools available. Eighteen percent of rural households said they did not send their children to preprimary school because they preferred to keep them at home, as did 54 percent of households from urban areas, where more households have the option of sending their children to preprimary school. Twenty-one percent of urban households say that there are no preprimary schools available to them (Figure 6.5). Figure 6.5: Reason for Not Attending Preprimary School in 2006/07 (percent) 70 66 60 54 50 40 30 21 18 20 13 11 8 10 4 4 2 0 Urban Rural None available/Too far Prefer to keep at home Too young Other Too expensive Source: TLSS 2007 and authors' calculations. Tuition is a barrier, but many parents prefer to keep their children at home. The most common reason for urban households not sending their child to preprimary school is that they prefer to keep their children at home. The cost of preprimary school is only the fourth-most-common reason why parents do not enroll their children. At 8 percent among urban households and 2 percent among rural households, the cost of preprimary school does not appear to be as large a barrier to access as might be thought. Lack of parental knowledge about the potential benefits of preprimary education and the fact that existing programs in preprimary schools do not help develop the skills of children could be the major reasons for the low level of interest in preprimary schools. A survey of the needs of children and mothers, conducted in 2004 by the MOE and the United Nations Children's Fund (UNICEF), showed that many parents (mothers) had limited knowledge about the needs of children at an early age. The TLSS 2007 also indicates that parents do not know when children are eligible for preprimary school. Eleven percent of parents think that their children are too young to enter preprimary school even though the children are within the normal age for it. Further, it is also known that preprimary schools usually do not prepare children to enter school, but function more as daycare. Parents might therefore also keep their children out of preprimary school since they see limited advantages to it. 65 The TLSS asked communities if they had a kindergarten in their community. The enrollment rate given here is the share of preprimary children in the communities with a kindergarten that were enrolled. 76 Hence, curriculum reform combined with parental education and advocacy could greatly improve the impact of preschool learning at little cost. Eventually, availability of institutions would also have to be addressed, but with little effective demand, little proven impact, and limited resources, this would not be a first priority. In addition to the factors outlined above, the relative absence of preprimary learning is also due to the fact that a culture of the preschool learning environment is not yet integrated into a society in which mothers prefer to stay at home with their children. Therefore, other approaches that would be relatively inexpensive and faster to implement than building institutions should have first priority. This could include advocacy campaigns on the importance of preschool learning, both at preprimary institutions and at home with the parents. Encouraging nongovernmental organizations and other private organizations to mobilize communities to provide preschool learning would also be a way for the Government to facilitate provision at little cost. Ensuring quality and standards would then be the responsibility of the Government. C.2 General education enrollment Tajikistan is approaching universal primary and lower secondary education but still needs to target those left behind. In 2007, 99.5 percent of boys and 98.6 percent of girls aged 8­10 were in school, while 98.4 percent of boys and 95.2 percent girls aged 11­15 were in school (Figure 6.6). Since enrollment is relatively close to being universal, a targeted approach for those left behind is needed, because broad programs for this purpose are likely to be inefficient.66 A targeted program is already in the planning stages using EFA/FTI catalytic funds. Differences in enrollment between wealth groups are modest but rising for both primary and secondary education. Substantial gender differences also persist. Inequality in enrollment based on consumption or income is lower in Commonwealth of Independent States (CIS) countries than in most parts of the world. However, among CIS and ECA countries, enrollment inequality across consumption in Tajikistan is one of the highest.67 This is true for both primary and secondary education (Asad, Murthi, and Yemtsov 2005). Further, these inequalities seem to be increasing over time. The primary enrollment ratio between the highest and the lowest consumption quintiles has increased from 1.05 in 1999 to 1.07 in 2007. For secondary enrollment, it went from 1.09 in 1999 to 1.24 in 2007 (Figure 6.6). Tajikistan also has some of the highest gender enrollment inequalities in the ECA Region. Further, between 2002 and 2005, there was no systematic improvement in gender inequality for either primary or secondary education (World Development Indicators). 66 The TLSS 2007 can tell us something about those out of school at this level, but it is not an ideal instrument considering the few students out of school. The standard errors on a survey are too large to reveal something meaningful about so few students. 67 Enrollment inequality across consumption and gender is often measured as the ratio between the enrollment rates in the highest and lowest quintiles, and the ratio between the enrollment rates for males and females. Here, these measures are also used to look at inequality in enrollment. 77 Figure 6.6: Top-bottom Consumption Quintile Enrolment Ratio, 1999­2007 1.25 1.2 1.15 1.1 1.05 1 Primary (age 7-14) Secondary (age 15-17) 1999 2003 2007 Source: Asad, Murthi, and Yemtsov 2005 and TLSS 2007and authors' calculations. Enrollment levels drop significantly after lower secondary education, and among those no longer in school are typically girls, poor, and those living outside Dushanbe. For both girls and boys, enrollment is strongly linked to age. At age 15, 94 percent on average are enrolled, while only 66 percent are enrolled at age 17 and 37 percent at age 18 (Annex 13, Table A13.3). The sharp drop in enrollment corresponds to completion of general secondary education. There is a clear pattern of lower enrollment for girls than boys across all ages. On average, in the age 8­18 group, 95 percent of boys are enrolled compared to 87 percent of girls (Annex 13, Table A13.3). The gender bias is smaller in Sogd and Gorno-Badakhshan Autonomous Oblast (GBAO), and among ethnic Uzbek compared to ethnic Tajik. Lower enrollment for girls is not concentrated in certain consumption quintiles since enrollment in all quintiles is lower for girls than for boys. However, enrollment is correlated with poverty and there is lower enrollment in the lower quintiles for both genders (Figure 6.7). Figure 6.7: Enrolment Rates across Gender and Consumption Quintiles 1 1 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0 8 9 10 11 12 13 14 15 16 17 18 8 9 10 11 12 13 14 15 16 17 18 Age Age Male Female Poorest Richest Source: TLSS 2007 and authors' calculations. 78 Poverty and tuition fees do hinder students from staying in school. Among the children not in school, most are "voluntary" out of school, but some children were not attending due to financial barriers (7 percent), poor facilities (5 percent), and school being too far (4 percent). Breaking down the reasons by location, one can see that there is a higher share of young people not being in school due to financial problems in areas where school fees are more common. As seen in the analysis of private expenditure on education earlier in this chapter, there are more schools that charge fees for general education in Dushanbe than other urban areas. This is reflected in 17 percent saying that they were not enrolled due to financial reasons in Dushanbe compared to 11 percent and 6 percent in other urban and rural areas. Community-level data also show that communities that have a high share of students paying tuition fees also have the lowest enrollment rates--a clear indication that tuition fees bar some students from staying in school. C.3 Higher education enrollment Enrollment in higher education increased from 2002 to 2006. The gross enrollment rate in tertiary education increased from 14 percent to 18.6 percent over four years. This enrollment level is low compared to other countries in the ECA Region but is significantly better compared to other countries with Tajikistan's development level. Higher education is largely a privilege of the more wealthy and of boys, resulting in very high gender inequality. The gender bias is most severe at the tertiary level, with a gross enrollment rate of 27 percent for males compared to only 10 percent for females in 2005 (World Development Indicators). This results in a female/male gender enrollment ratio of 0.35, which is among the lowest in the region (Figure 6.8). A number of countries also have a strong gender bias in tertiary education, but most have a bias against male students since females make up the majority of the student body. This can be seen from the average female/male gender ratio of 1.26 compared to Tajikistan's 0.35, with 1 being equal enrollment. Figure 6.8: Gender Inequality in Tertiary Education, 2005 2.00 Gender equal 1.75 enrollment 1.50 1.25 1.00 0.75 0.50 0.25 ia on R M F B blic C lic k om ic zb u n eo n R kr ia un n rb an Rulg ia lb a u ia st ia ubia ua a h ia A ki ey La ia az v d va R uble on atis rm ia L ldo y ia Y U T ta G ija A ni H sta ith v in u o r K lo n z U en E an g ak en epan on r tv A at ,F M ga e rk ep a ze st ed er r S a is ch B r epa a a ro ol jik el P ac ed R Ta gy an ze lo yr si C S K us R Note: The figure shows the ratio of female-to-male enrollment in tertiary education. A ratio of 100 is perfectly equal enrollment. The further the distance from 100 the greater the inequality. Source: World Development Indicators. Transparency of access to and quality of high education are likely to improve with upcoming standardized entrance examinations. There are plans to introduce common standardized examinations that can be used for all higher education institutions. This will be a first important step in allowing those with the greatest potential to access higher education. Poverty and tuition fees seem to bar some people from entering higher education. Some young people say they are not enrolled in higher education due to financial challenges. This is particularly true in 79 Dushanbe and among males--groups that are most likely to enter higher education. Analysis of community data reveals there are higher enrollment rates in areas with no tuition fees. There are also high private costs associated with higher education (Annex 13). Finance reform and more student finance options are needed to ensure access, quality, and financial viability. Current block financing of universities provides no incentives for universities to adapt curriculum, staff, and teaching to the challenges of today and tomorrow. In other countries, competitive funds have been shown to improve incentives. Further, if subsidies were at least partly based on number of students, higher education institutions would have more incentive to adapt in order to attract more students, who in turn would request education relevant to the labor market. A per-student finance system should also end the current arbitrary system where roughly half the students study for free while the other half has to pay. To accommodate and attract students from a wider social background than just the better off, access must also be improved. As a starting point, the already existing scholarships should to a larger degree benefit those most in need. In the longer term, student loans are also an instrument that can not only increase enrollment, but also keep government expenditure on higher education in check. Expanding enrollment in higher education can be very expensive if all or most costs are borne by the Government. Hence, it is important that the costs are moved to those that will benefit from the education.68 D. Indicators of School Quality An environment conducive to learning depends on many aspects, including a school being in reasonable proximity to the student, and having good facilities, qualified teachers offering relevant education, and no barriers to entry for the students. This section will look at some of these aspects, such as the quality of the school infrastructure evaluated by local residents, availability of teachers, and student access and absenteeism. D.1 School facilities School facilities are of generally low and unequal quality with worse facilities in poor and rural areas. Lack of satisfactory physical facilities such as buildings, desks and chairs, and blackboards, is mostly a rural problem, with poorer areas having worse facilities. For instance, rural communities representing 40 percent of the population report that their schools do not have satisfactory buildings, desks and chairs, and blackboards compared to about 10 percent of the urban population reporting the same problems. Over 60 percent of students living in the poorest communities do not have satisfactory textbooks and heating fuel. These problems are also widespread in relatively richer communities. By joining all the indicators into a standardized index and running local regressions against the average consumption in each village (a Lowess regression) we can show the clear and smoothed relationship between unsatisfactory facilities and low welfare status (Figure 6.9). 68 See, for instance, Johnstone 2000 and 2004, for more discussion on the best practice and experience from other countries on the subject. 80 Figure 6.9: Quality of School Facilities and Average Consumption in Community d x f is a fa tio ith c o l c s In e o d s tis c nw s h o fa ilitie 2 1 0 -1 50 100 150 200 250 300 avc Average consumption in community bandwidth = .8 Note: Each observation shows average consumption in a community and an index of dissatisfaction with school facilities in the community. A community is defined as a Primary Sampling Unit in the TLSS 2007 survey. The line is a smoothed Lowess regression line (estimated by running local regressions across all values). Source: TLSS 2007 and authors' calculations. However, school infrastructure generally improved from 2002 to 2007. Over 50 percent of communities report improvement in the last five years in key aspects of school facilities such as buildings, desks and chairs, blackboards, textbooks, and heating fuel. On the other hand, up to 17 percent of communities report that some of these facilities have deteriorated over the last five years, with the remaining reporting no change. Most progress is reported for buildings, where 65 percent of communities report improvement. Classroom basics such as desk and chairs, blackboards, and textbooks have also improved in 57 percent of the communities, while progress in heating fuel is reported in only 48 percent of the communities. Most improvements were concentrated in rich and urban areas, while deterioration was more common in rural and poor areas. Across all type of facilities, rich communities were more likely to report improvement than poorer communities. For instance, among the poorest quintile of communities, 46 percent reported that their textbook situation had improved over the last five years, while 59 percent among the richest quintile of communities reported such improvement. In line with urban areas generally being wealthier than rural areas, urban areas were also more likely to report improvement in infrastructure than rural areas. The neglect of schools in rural areas compared to urban areas can also be seen in the higher share of communities in these areas that report deterioration in school facilities compared to urban areas. Areas that reported the greatest need for school improvements were the least likely to report recent improvements. Rich and urban schools have the best quality, while these were also the schools that reported the most progress in quality. The higher quality in urban and rich areas could be due to the recent improvements, but it could also indicate that the improvements did not take place where most needed. In any case, it shows that future improvements in quality have to take place in poor and rural areas. School closures for a certain period are a significant problem, particularly in Sogd. Twenty percent of all schools were closed for at least one week during the 2006/07 school year. Again, the extent of the problem varies greatly across oblasts, with 40 percent of schools in Sogd being closed--compared to 18 percent in the Region of Republican Subordination (RRS), (10 percent in Khatlon, and 2 percent in Dushanbe). The schools that were closed for one reason or another were on average closed for seven weeks. The reasons for school closure vary by oblast. In Sogd, schools were closed mostly due to bad 81 weather, while in Khatlon they were closed due to agricultural work. In general, bad weather (combined with an unsatisfactory supply of electricity and heating) and engagement of pupils in agricultural work are the two factors that kept schools closed most weeks (56 percent combined). D.2 Access (distance and time) to schools Access to general schools is relatively good but seems a greater challenge for girls. The relatively good access is exemplified by the fact that 98 percent of students in general education live within 5 kilometers of the school and 95 percent have a commute of 15 minutes or less. The distance to schools increases with higher levels of education, but 63 percent of students at the secondary specialized level and 55 percent in higher education still live within 5 kilometers of the school they are attending. Ninety-three percent of all students commute less than 15 minutes one way and only 1 percent commute more than 45 minutes one way. Relatively few parents who have children of the general school age report that long distance to school is the reason why their children are not attending school (only 4 percent of all non- attending children are due to this reason). However, more girls than boys are not attending school due to distance--5 percent of girls compared to 1 percent of boys. Student absenteeism is a limited problem. Around 20 percent of students missed at least one week of class in 2007, but less than 6 percent missed more than two weeks. This indicates that students rarely missed many weeks of school. Most of those that report that they missed more than two weeks of classes did so due to their own illness (49 percent) or bad weather (18 percent). Work and in particular agricultural work also keep some students away from school. This is a particular problem in Sogd and Khatlon where, respectively, 17 percent and 27 percent of the students that missed two weeks or more did so because they had to work. D.3 Teachers Schools and communities consistently report a shortage of qualified teachers, particularly in rural areas. Thirty-eight percent live in communities that report having unfilled teaching positions. Teacher shortages are reported to be much higher in rural areas (44 percent), and particularly in the RRS (69 percent). The lack of teachers seems more related to location than to the overall wealth of the location. Remoteness and severe living conditions could be factors influencing the many unfilled positions. Meanwhile, the student/teacher ratios are low compared to other countries at the same development level, indicating that there is generally a sufficient number of teachers available. The student/teacher ratio at primary and secondary levels was, respectively, 22.2 and 16.5, in 2006. Among small schools (less than 150 students) the average student/teacher ratio across the primary and secondary levels is 10, and among schools with 150 to 225 students the average is 14 (World Bank 2008b). This puts the Tajikistan student/teacher ratios much lower than other countries at a similar development level (UNESCO education database). An educational front-runner, like the Republic of Korea, for instance, has primary and secondary student/teacher ratios of 28 and 18, respectively, while most countries with a GDP per capita similar to Tajikistan are in the 30 to 50 range for primary and 20 to 35 for secondary. Institutional challenges emanating from teacher and school regulations might hinder optimal use of available teachers. The reported lack of teachers combined with relatively low student/teacher ratios points to institutional challenges in using available teachers efficiently. Experience in Azerbaijan, for instance, showed that a large number of mandatory subjects, combined with a high degree of specialization and inflexible teaching practices, resulted in a very high number of required teachers per school. A school that offered all mandatory subjects for grades 1­11 would require 31 teachers to provide these classes. In a school with 15 students per grade, this would equal a student/teacher ratio of 6. An in- 82 depth study is needed to fully understand the situation in Tajikistan, but there are strong indications that institutional changes such as providing more flexibility to schools in establishing a balance between mandatory and optional subjects, and having teachers trained to teach several subjects and grades, could lead to the optimization of the number of teachers while increasing the teaching workload and salaries. Other options such as merger (consolidation) of schools and incentive pay for rural and multi-grade and multi-subject teachers could also be considered options for alleviating the perceived demand for teachers. E. Conclusions There has been progress in the education sector in Tajikistan, but the country still faces major challenges in improving education outcomes. Tajikistan has managed to sustain high enrollment ratios in primary and secondary education. Both public and private funding for education also increased significantly to support this trend. However, many challenges remain, including: (a) deficient school infrastructure, including access to electricity and heating in winter months (resulting in temporary school closures); (b) a perceived lack of qualified teachers especially in rural areas, while student/teacher ratios are comparatively low; (c) the hidden costs and/or tuition fees representing a substantial burden for poor households; and (d) public financing arrangements rely significantly on the budgets of local authorities, which introduces inequalities in the quality of education services across regions. The quality of school facilities is unequal and often very bad. Rural and poor schools are particularly affected by poor infrastructure, and they saw the least improvement in facilities over the last five years. Over 40 percent of rural areas report their school facilities to be of unsatisfactory quality in all aspects. At the same time, there is evidence that improvements might not have taken place in areas that have the greatest need. There is a reported lack of qualified teachers; however, at the same time, there are comparatively low student/teacher ratios, which points to institutional challenges. The lack of teachers seems to be more related to location specifics and low pay rather than to the overall wealth of the location. At the same time, the student/teacher ratios are comparatively low, indicating that the major challenge is to adjust the institutional setup to use and allocate teachers optimally. Adjustment of curriculum, promotion of multi-subject and multi-grade teachers, potentially combined with payment incentives for rural schools, should be looked into further. The use of tuition fees or hidden charges in general education represents a significant problem for poor households. Analysis indicates that 13 percent of children in public general education pay tuition fees or charges, despite the free school policy. The practice is most common in Dushanbe (49 percent). There is a clear link between areas with school fees, lower enrollment, and children saying they are not attending school for financial reasons. Enforcing free education for all is important to achieving universal general education. Most schools are locally financed and are therefore dependent on available funds and the decisions of local governments, which limited central government oversight. This system results in unequal funding of schools, hinders federal oversight, and provides no incentives for efficiency at the school level. Current reforms are addressing some of these issues. Per capita funding, budget reforms, and more school autonomy are likely to contribute to better oversight and efficiency. The impact of those reforms needs to be assessed as more data and evidence become available. 83 Chapter 7: Health and Poverty A. Introduction Evidence from various countries shows that the poor have poorer health. They have more episodes of morbidity and suffer from higher death rates than the less poor. Despite having poorer health, they have less access to health care services and spend a higher proportion of their income on health care. A health system that does not consider the poor can exacerbate the difficulties that the poor already have in accessing and using health care services. Bad health can lead to poverty when an illness prevents full participation in the labor market. In Tajikistan, the gap in health outcomes between rich and poor is clearly shown by much higher rates of child morbidity and mortality in poor households. Data from the Multiple Indicator Cluster Survey (MICS) 2005 show that under-5 mortality rates were 100 per 1,000 live births for the poorest 60 percent of households compared to 74 per 1,000 live births for the richest 40 percent. The 2007 Tajikistan Living Standards Survey shows that almost 46 percent of the poorest households found it impossible or very difficult to pay for health care compared to 28 percent of the richest households. The analysis showed that the deterioration of the health of the household head is linked to a 15.6 percentage point increase in the probability of becoming poor. Reasons for poor health outcomes in Tajikistan, like in other societies, are complex. Health outcomes are determined by an interaction of numerous social and economic factors. Despite some factors clearly being outside the control of the Ministry of Health (MOH), the health system can be redirected toward more pro-poor health care provision, thus making the differences in health outcomes between the poor and non-poor less pronounced. This chapter analyzes data on the relationship between poverty and health from the 2007 Tajikistan Living Standards Survey (TLSS 2007), the most recent survey. Background information from recent publications is used to provide context. Where data allow, the chapter presents changes in poverty and health patterns since the previous survey and previous poverty assessments. Despite the Government's increased allocations to the health sector, Tajikistan continues to have poor health outcomes, with the poor population suffering more from the lack of adequate access to health services. To improve equity in health service provision, key health financing reforms must be prioritized. Without financial protection, high out-of-pocket expenses for major episodes of illness can also impoverish non-poor households. B. Overview of the Health System The Tajikistan health system leans toward serving the better-off more than the poor. Tajikistan's health system is largely public and is partly financed by funds raised through general revenues. For many years, every citizen was entitled to free health care, but because of meager budget allocations--­only 17.7 percent of total health expenditure is publicly funded--the amount allocated to the sector covers only minimal needs. Foreign aid is an important source of financing.69 Surveys indicate that through formal 69 Foreign aid represents a substantial form of financing; there are a large number of nongovernmental organizations providing targeted interventions in rural but also poor urban populations. Donors support most of the activities related to reforms in the health sector, and vertical programs such as immunization and HIV/AIDS. Almost all have units outside the MOH that have 84 user fees and informal out-of-pocket payments, households contribute about 72.4 percent of total health expenditure. Tajikistan's system of health care financing has even worse implications for the poor, because it does not offer risk protection, nor is it equitable. The poor that become sick usually go untreated or fall further into poverty. The state is the main provider of health care services. Health care services are divided into four horizontal levels and separate vertical pillars for the national program (European Observatory 2000). Box 7.1 describes the organization of the health system. Box 7.1: Organization of the Tajikistan Health System The Ministry of Health (MOH) is responsible for identifying priorities and developing health policy, disease control, coordination and management of federal-level institutions, research institutes, and health educational institutions for health professionals. It regulates and manages national-level health facilities. Oblasts (regions) and rayons (districts) run most regional and local health services including general hospitals, and primary health care (PHC) and public health services. The Ministry of Finance (MOF) determines the overall rayon budget and sector-specific ceilings. Local budgets are formulated at the rayon level through the collection of local taxes, and local budget deficits are covered by the Ministry of Finance (MOF) through subsidies from the national budget. Each health facility formulates its own annual budget request, which are aggregated by financial departments at the rayon level and then sent to oblast financial departments, which aggregate the overall oblast health budget and send it to the MOF for approval. The MOH is responsible for consolidating and negotiating the budget with the MOF for federal health facilities. Oblast health departments have dual accountability--to the MOH on professional matters and to the oblast administration on financial issues. The Chief Physician of the Central Rayon Hospital has deputies responsible for rural clinics, polyclinics, disease prevention, and mother and child services. The central management of most health services is located within hospitals. Rayons are also responsible for defending their budgets to the MOF and have flexibility and discretion in fund allocation among agencies. In 2007, the health system employed 13,392 physicians and 29,662 midlevel personnel that included 21,142 nurses. The sector has a large infrastructure of buildings: over 2,300 PHC facilities, 335 hospitals of different types, 112 dispensaries, and more than 300 facilities related to the public health subsector. Tajikistan inherited a centralized, large, and highly specialized health system that emphasizes curative and inpatient care over primary care. Due to scarce resources, a large proportion of hospitals and health centers are deteriorated and in need of repair. Medical equipment in these buildings is in need of updating, and shortages of drugs and supplies are common. A recent fiduciary assessment concluded that the capacity of the health staff in financial and procurement management is weak, and that the multiplicity of actors and the general lack of information between providers and consumers creates an environment of opaque governance practices. Public expenditures continue to be dominated by number of hospital beds and staff. The large supply of hospitals consumes the largest chunk of the budget. Tajikistan has a higher number of hospitals and beds per capita, low inpatient admission rates, and long hospital stays compared to countries with more efficient health care systems (ADB 2007). In 2007, 65 percent of total public health expenditures project management functions. The Bank supports the health sector through the Community and Basic Health Project, a Programmatic Development Policy Grant, the Avian Influenza Project, and the Central Asia AIDS Project. Key partners in the sector are the Swedish International Development Agency (SIDA) and the Swiss Agency for Development and Cooperation (SDC). Other partners include the United States Agency for International Development (USAID), the Asian Development Bank, the Aga Khan Foundation, and the Global Fund, which funds the national programs on HIV/AIDS, tuberculosis (TB), and malaria. 85 was spent on hospitals, and the rest was distributed to polyclinics, public health, and miscellaneous categories. This has created an under-resourced and poor-quality primary health care system, with about 80 percent of patients bypassing primary health care centers to seek care at the next level. Tajikistan's health system offers few incentives for delivery of quality services. Most health sector workers and health care providers are public employees receiving an average monthly wage of 119 somoni for primary health care (PHC) staff and 78 somoni for secondary health care staff (equal, respectively, to the US$ 27 and US$ 18 per month), which is significantly lower than the average wage in the country. The budget for health care providers is not based on performance, and since management is centralized, hospital managers have no control over budgets and have no incentives to offer more cost- effective services. To supplement their pay, about 18.5 percent of the health workforce has second jobs outside their usual facility, working on average an additional 21 hours per week in those jobs (Dabalen and Wane 2008). Almost half of health care workers receive informal payments to supplement their low incomes. According to Dabalen and Wane (2008), the average health care worker levies approximately 28.8 somonis per month. Doctors charge almost twice as much. Further, 46 percent of health care workers admit to receiving informal payments gifts, in cash and in-kind. TLSS 2007 data show that only 17 percent of those who paid for care in the last four weeks received a receipt for their payment. Reforms were recently introduced to improve efficiency, access, and quality of care. These include strengthening primary health care by improving its management and introducing per capita financing. In June 2007, the MOH introduced a Basic Benefits Package (BBP) of hospital health services in four pilot rayons. The BBP aims to provide more equitable access to hospital services and at the same time represents a first step in moving from input- to output-based financing. The BBP provides free services for vulnerable population groups and provides a legal framework to formalize informal payments by introducing a simple co-payment mechanism for the rest of population. According to the MOH, piloting of the BBP has yielded some positive results including a decrease in levels of out-of-pocket payments. However, due to the current low level of state funding for the sector and excess capacity at the hospital level, the proceeds generated by co-payments70 do not adequately cover the cost of the basic package of services. C. Health, Nutrition, and Population Status Since the 1970s, Tajikistan has experienced rapid fertility declines, but fertility rates still remain very high. During the Soviet era, the population in Tajikistan grew faster than any other Soviet republic, with an annual growth rate of 3 percent in the early-to-mid-1970s. The current growth rate is about 1.5 percent as a result of declines in fertility, which has dropped from 6 children per woman in the 1970s, to 3.8 children per woman in 2005. It is projected that by 2030­35, fertility will have declined to the replacement level of about 2.1 children per woman (United Nations 2007). Due to high fertility rates, Tajikistan has the highest dependency ratio in the Europe and Central Asia Region (Figure 7.1). 70 In 2007, the co-payment was 30 percent and 70 percent of the total cost of a service for, respectively, patients with a referral for PHC and for patients with no referral. In 2008, in order to make BBP more financially sustainable, the MOH adjusted the co- payment structure to 50 percent and 80 percent, respectively. 86 Figure 7.1: Population Dynamics, 2000­40 T o t a l D e p e n d e n c y R a t io s 9 0 8 0 7 0 6 0 K a z a k h s ta n 5 0 K y r g y z s t a n T a ji k i s t a n T u r k m e n s it a n 4 0 R u s s ia 3 0 2 0 1 0 0 2 0 0 0 2 0 0 5 2 0 1 0 2 0 1 5 2 0 2 0 2 0 2 5 2 0 3 0 2 0 3 5 2 0 4 0 M e d ia n A g e s 5 0 4 5 4 0 3 5 3 0 2 0 0 5 Years 2 0 1 5 2 5 2 0 2 5 2 0 3 5 2 0 1 5 1 0 5 0 K a z a k h s ta n K y r g y z s ta n T a j ik is t a n T u r k m e n is t a n R u s s ia Source: United Nations World Population Prospects (2007 revisions). Tajikistan's median age, at about 20, is one of the lowest in the ECA Region and is projected to increase to about 26 by 2025. In comparison, Russia currently has a median age of about 37, which is projected to increase to 42 by 2025 (Figure 7.1). Since fertility transition started much later than in other ECA countries, about 62 percent of the population was in the 0­24 age group in 2005, and as the effect of the fertility transition continues, this proportion will drop to 45.2 percent by 2030. The proportion of the population of working and reproductive age is expected to remain large--those aged 15­64 comprised about 57 percent of the population in 2005. This figure is expected to increase to 66 percent in 2030. This large proportion of the population that is young combined with declining fertility rates means that total dependency ratios will remain favorable for economic growth for future decades. As has been found in other surveys, the majority of respondents in the 2007 survey identify health as their issue of greatest concern, compared to 24 percent who cite money or jobs. Around half of all households in 2007 cited health as their major concern. Though most respondents identified health as a major concern, there are differences by poverty status, with a higher proportion of respondents in less- poor households mentioning health as a concern compared to individuals in poorer households (Figure 7.2). 87 Figure 7.2: Aspects of Life Concerns by Household Consumption Quintile Figure 1: What aspect of life concerns you most at present? By household consumption quintiles 60 Percent Respondents 50 40 30 20 10 0 Money Job security Health Safety Other Q1 (poorest) Q2 Q3 Q4 Q5 Source: World Bank estimates using TLSS 2007 data. Health indicators in Tajikistan are among the lowest in the ECA Region. While average life expectancy at birth is comparable to neighboring countries, Infant Mortality Rates (IMRs), under-5 mortality rates, and the Maternal Mortality Ratio (MMR) are higher (UNDP 2005; UNICEF 2007) (Table 7.1). Like other countries in the region, Tajikistan faces a burden of non-communicable diseases that is higher than that of communicable diseases. According to the WHO, chronic illnesses accounted for 67 percent of all deaths. Table 7.1: Health Outcomes in Tajikistan in Compared to Neighboring Countries Country Gross National Life Expectancy Infant Under-5 Maternal Income Per 2006 Mortality Rate Mortality Rate Mortality Capita 2007 Rate 200 2000 2000 2006 2005 6 Tajikistan 460 67 75 56 93 68 97 Kyrgyzstan 590 68 44 36 51 41 150 Uzbekistan 730 67 52 38 63 44 24 Kazakhstan 5060 65 37 26 43 29 140 Russia 7560 66 16 10 20 13 28 Ukraine 2550 68 19 20 24 24 18 Source: World Bank Health Nutrition Program Statistics and 2005 MICS. Tajikistan faces significant challenges in meeting the Millennium Development Goals (MDGs). According to the United Nations Development Programme's (UNDP's) MDGs Needs Assessment for Tajikistan, the total cost of meeting health MDGs is estimated at about US$3.6 billion (US$46 per capita) during 2005­15. Tajikistan still needs to make a significant effort to achieve targets for MDGs #4 (reducing child mortality) and #5 (improving maternal health) (UNDP 2005; World Bank 2005), and targets on TB, HIV, and access to safe water, which is a major source of morbidity and mortality for children. Data show that morbidity and mortality among children are closely linked to poverty. Health outcomes among infants and children are associated with lower socioeconomic status of their mothers and the households they live in. For example, episodes of diarrhea, a major killer of children, are higher among children from poorer families. Data from the 2005 MICS show that 18 percent of children in the poorest wealth category had diarrhea in the last two weeks before the survey compared to 11.6 percent of 88 children in the richest category. Under-5 mortality rates were 100 per 1,000 for the poorest 60 percent compared to 74 per 1,000 for the richest 40 percent. Moreover, when children from poorer households became ill, their illnesses were less well managed. The estimated number of people living with HIV/AIDS in Turkistan is growing. Estimates from the United Nations Joint Programme on HIV/AIDS (UNAIDS) suggest that there were 10,000 people living with HIV/AIDS at the end of 2007 (range 5, 000 to 23, 000). This is a large increase from 2,500 people in 2001, and the prevalence level is much higher than the number of officially reported cases, which is 738 (UNAIDS 2008). The major mode of transmission continues to be intravenous drug use, which is fueled by the drug trade in the region. The drug trade complicates the fight against HIV/AIDS. Although access to information is growing, a large proportion of the population has never heard of HIV. TLSS 2007 data show that as many as 32 percent of respondents had not heard of HIV/AIDS. Eighty-seven percent of those that had not heard of HIV/AIDS were rural residents and only 2.7 percent were residents of Dushanbe. Ninety-five percent lived in Sogd, Khatlon, and the Region of Republican Subordination (RRS). Poor people are less informed about HIV (Figure 7.3). Figure 7.3: Variation in the Proportion of the Population that Has Heard of HIV/AIDS 10 0 90 80 70 60 50 S e ri e s 1 40 30 20 10 0 l or n or ia n r r io o Q oo 1Q 2Q 3Q 4Q io ec po po po at 5 at lp p uc on n uc n S ra no a d ed ln ry rb ru rE n da ra U a o he N ru rb on ig U ec H S Source: TLSS 2007. Analysis of TLSS 2007 data shows that among women of reproductive age (15­49), 41 percent were using contraceptives, and overall there appears to have been little change in the contraceptive prevalence rate since 1999. Almost 37 percent of women aged 15­49 who were married or in union at the time of the survey were using some method of contraception. Variations by urban/rural residence were not as large as expected, although there were quite large variations by oblast, ranging from 55 percent in Gorno-Badakhshan Autonomous Oblast (GBAO) to only 33 percent in Khatlon. Miscarriage/stillbirth rates are still high and are probably a proxy for a number of factors that affect maternal and child mortality. Over 17 percent of all women aged 16­49 who had ever been pregnant reported that they had had at least one miscarriage/stillbirth. The TLSS 2007 data show that the rural and poor have higher rates of malnutrition. Data show that, for example, about 31 percent of children in urban areas were either moderately or severely stunted compared to 42 percent in rural areas. The proportion of children either moderately or severely stunted 89 was 30 percent in Dushanbe compared to 38 percent in Sogd, the RRS, and GBAO, and 43 percent in Khatlon. D. Access to and Utilization of Health Care Analysis of utilization rates shows that the poor have many fewer visits than the less poor. The 2007 data confirm the pattern that was observed in the 2003 data--that the poor have fewer outpatient visits compared to the less poor. However, the 2007 data show that the number of outpatient visits has increased. The data also show that the difference in the average number of visits per year between the poorest and the highest consumption quintiles has narrowed. Access to and utilization of health services is also likely to be determined by availability of services in the various geographic localities. For example, Khatlon, while being one of the poorest regions, also has one of the lowest numbers of physicians and nurses per 1,000 population (Figure 7.4). Figure 7.4: Outpatient Visits per Year, 2003 and 2007 Outpatient Visits Per Year, 2003 and 2007 1.40 1.20 e a ita 1.00 isits P r C p 0.80 0.60 0.40 V 0.20 0.00 1st 3d 5th 2003 (poorest) (highest) 2007 Household Consumption Quintile Group Source: TLSS 2003 and 2007. The TLSS 2007 data show that a significant share of households have family members who delay seeking help or did not seek help at all for financial reasons. About 45 percent of those in the poorest quintile reported finding it impossible or difficult to pay for health care compared to about 29 percent in the richest quintile. The majority of those who have difficulty finding finances for health care are either rural residents or the poor, including the urban poor. However, the population that did not receive medical care due to lack of money decreased from 50 percent in 2003 to 33 percent in 2007. This most likely reflects a decreasing level of poverty during the period. The TLSS 2007 data also show that a significant share of households had family members who delayed seeking help or did not seek help at all three or more times in the last 12 months. The majority of those who did not access care even when they had been referred to the hospital were in the lowest consumption quintiles who resided in rural areas (Figure 7.5). Seventy-six percent of these 71 respondents lived in Khatlon and the RRS, the majority of whom were rural residents. More than 54 percent of respondents said that in the last 12 months someone in their family had been referred to the 71 Khatlon and the RRS combined account for 58 percent of the total population (mostly rural) and 52 percent of the total poor in Tajikistan. 90 hospital but did not go. Lack of finances was leading people not to access care, even when they had been referred to the hospital for treatment. Lack of finances was mentioned by slightly more than 50 percent of respondents. Figure 7.5: Distribution of Respondents by whether Finding Money for Care Was Impossible, Difficult, or Not Difficult 50 45 40 35 30 I m p o s s i b le / d i f f ic u l t 25 N o t D i f fi c u lt 20 15 10 5 0 r r or r oo oo an n al oo Q Q Q Q Q a po ur -p t -p rb 5 3 4 2 1 p is R on n ik al an U no aj ur ln rb T R an ra U rb u R U Source: TLSS 2007. Utilization of maternal and child health services varies by location and poverty level. In general, 88 percent of the women reported having had at least one medical consultation during their last pregnancy, but there are significant differences by location and socioeconomic status (Figure 7.6). Figure 7.6: Proportion of Women that Had at least One Antenatal Visit during Last Pregnancy 10 0 95 90 85 199 9 80 200 3 200 7 75 70 65 60 il e il e e n O LL al an gd S nb tl o R nt nt ur BA A rb Su R ha ha ui ui R U G q tq us K st es D re h oo ic R P Source: TLSS 2007. 91 About a quarter of the women in Tajikistan delivered their children at home. This seems to be more of a rural phenomenon and not only a poverty issue, since rural non-poor also mostly delivered their children at home. What is encouraging is that the TLSS 2007 data suggest that home deliveries are declining and deliveries attended by a medically trained person are rising compared to TLSS 1999 data. Vaccination coverage remains strong. Almost all respondents (90 percent) mentioned that children under 1 year of age in their community had been vaccinated, and about 86 percent responded that there had been an immunization campaign in their community in the last year. The differences by income and, hence, poverty profile, were minimal. In addition to financing constraints, the poor might be deterred from visiting health care facilities because of drug shortages and poor quality of services offered. The 2007 Public Expenditure Tracking Survey report concluded that only 61 percent of facilities surveyed reported receiving funds for drugs or actual supplies of drugs. A significant share of facilities surveyed--16 percent--reported receiving no funds, except for payroll, or in-kind resources from the Government in 2005. In addition, only 50 percent of health care facilities had drugs in stock. Availability of electricity, heating, and so forth also remains a major issue, especially in rural areas and during winter months (Table 7.2). Table 7.2: Availability of Electrical, Heating, Water, Transport, and Communication Services in Health Facilities Urban Rural Tajikistan (Average no. of hours per day) Electricity Availability in Winter Period 19 6 9 Heating Availability in Winter Period 12 4 5 (% of facilities with access) Access to Piped Water 95 29 41 Access to Communications (phone or radio) 85 9 23 Access to Transportation Vehicles 48 12 19 Source: Tajikistan Health PETS 2006. E. Health Sector Management, Financing, and Expenditure Health sector management and financing can have huge implications for the health of the poor. The Tajikistan health system faces a number of challenges, including duplication and fragmentation of responsibilities, poor linkage between policy priorities and budgets, a multiplicity of actors, and lack of information flow between providers and users, which makes financial management a big challenge. The input-based budget formulation process perpetuates inequities in public spending on health. In 2007, public funding per capita ranged from US$3.3 in Khatlon oblast and RRS to US$7 in Dushanbe and US$10.5 in GBAO. Oblasts also allocate different proportions of their budgets to different subsectors of the system (Table 7.3). For example, GBAO spends 70 percent of its budget on hospitals compared to 46 percent in Dushanbe, which has the lowest expenditure on hospitals.72 72 Dushanbe hosts the majority of republican tertiary facilities that are financed by the federal budget. 92 Table 7.3: Structure of Public Expenditures on Health, 2007 (percent of oblast total health expenditure) Dushanb RRS Khatlon Sogd GBAO Tajikistan e Hospitals 47% 64% 66% 71% 76% 65% Polyclinics 37% 29% 22% 16% 12% 23% Public Health 5% 6% 9% 7% 8% 7% Other 12% 1% 3% 6% 5% 5% Health Expenditure by Local US$6.9 US$3.3 US$3.3 US$4.5 US$10.5 US$4.2 Budgets Per Capita US$ Source: Ministry of Finance and State Statistical Committee. The budget allocated to the health sector is increasing but remains insufficient to meet sector needs. About 15 years ago, Tajikistan spent 5 to 6 percent of its gross domestic product (GDP) on health, an amount that was considered sufficient. The economic collapse of the mid-1990s led to a decline in resources allocated to the sector. However, in the last several years the situation improved somewhat. Public spending on health increased from US$2.1 per capita in 2003 to US$6 in 2007. Yet, public spending on health was an estimated 1.2 percent in 2007.73 Most public expenditure on health comes from local government, which accounted for 76 percent of the total public expenditure in 2007. Central government health spending accounted for about 23 percent of total health expenditure (Table 7.4). The government allocation for health is used for paying salaries (47 percent); key inputs such as medicines, food, and fuel (41 percent); and repairs and maintenance (12 percent). In 2007, hospitals accounted for 65 percent of total public health expenditure, while polyclinics and public health services received 17.1 percent and 6.5 percent, respectively, a slight improvement in non-hospital subsector financing compared to 2003. Table 7.4: Comparison of Health Spending Patterns in Tajikistan, 1999, 2003, and 2007 1999 2003 2007 Total Health Expenditure (current US$ million) US$45.0 US$75.5 US$238.2 Population (million) 6.1 6.5 7.2 Per Capita GDP US$171 US$200 US$518 Per Capita Total Health Expenditure (US$) US$7.40 US$11.60 US$33.2 Total Health Expenditure, as % GDP 4.3 % 5.8 % 6.4% Government Health Expenditure, as % GDP 1.1 % 0.91 % 1.14% Household Expenditure on Health, as % GDP 2.9% 4.1% 4.6% Public Spending Per Capita (US$) US$1.7 US$2.1 US$6.0 (2001) Federal Health Spending (% of total) US$21 (2002) US$25 US$23 Local Health Spending (% of Total) 79 (2002) 75 77 Source: Health expenditure data are from Ministry of Finance and TLSS 2007. Population data are from the State Statistical Committee. 73 The nominal total health expenditure reached US$238 million and constituted 6.4 percent of GDP. As a proportion of all government expenditure, health accounted for only 4.15 percent in 2007, most of which was salaries. This is a decrease from 4.8 percent in 2002 and 5.2 percent of total budget expenditures in 2006. 93 Most spending on health is borne by households. Data from the TLSS 2007 confirm that households continue to spend a significant amount on health care (Table 7.5). Since the overall level of public funding continues to be low, it is estimated that, in 2007, the population made out-of-pocket expenditures (OOP) estimated at 4.6 percent of GDP, which is higher than the 3 percent of GDP spent in 2003. It is estimated that, in 2007, the OOP spending amounted to 72.4 percent of total health expenditures. However, the share of health care in household expenditure has been stable at around 5 percent in 1999 and 4 percent in 2007. Table 7.5: Distribution of Health Care Expenditures, by Sources and Service Categories, 2003 and 2007 (in percent) By Source % Total Health Government Household Donors Service Categories Spending Budget 2003 2007 2003 2007 2003 2007 2003 2007 1. Hospitals 33 32 32 33 42 59 26 8 - Drugs 11 9 13 6 52 89 35 4 2. Ambulatory care 64 60 3 5 90 89 7 6 - Drugs 52 44 1 0.3 96 98 3 1 3. Public Health 1 4 86 27 0 - 14 73 4. Administration 3 3 100 87 0 - 0 13 TOTAL 100% 100% 16.0% 17.7% 70.6% 72.4% 13.3% 9.9% Source: TLSS 2007 and Tajikistan Health Sector Note, 2005. Data suggest that the gap in health expenditure between the richest and poorest populations has become more pronounced in recent years. In 2003, the poorest quintile group spent 31 percent of the amount spent by the richest quintile group, whereas in 2007, this percentage decreased to 20 percent (Table 7.6). For 2007, the largest inequality is observed in the category of household expenditure at the primary health care level. This situation is attributed to both the lower utilization rates by low-income groups and financial constraints they face to purchasing necessary drugs. Table 7.6: Spending on Health by Sources, Lowest and Highest Consumption Quintile Groups, 2003 and 2007, in Million Somoni Total Spending Lowest Quintile Highest Quintile 2003 2007 2003 2007 Outpatient Services 15.3 29.9 45.5 256.1 - by Government 0.5 3.8 1.5 6.3 - by Households 13.7 21.7 40.7 242.2 - by Donors 1.1 4.5 3.3 7.6 Inpatient Services 7.6 37.3 28.8 72.8 - by Government 2.4 10.1 9.1 25.2 - by Households 3.2 24.6 12.1 41.1 - by Donors 2.0 2.6 7.6 6.4 Total Health Services 22.9 67.2 74.3 328.9 - by Government 2.9 13.9 10.6 31.6 - by Households 16.9 46.3 52.8 283.3 - by Donors 3.1 7.1 10.8 14.0 Source: TLSS 2003 and 2007. 94 F. Other Determinants of Health Outcomes: Distance to Health Care, Sources of Water and Sanitation and Nutrition Distance to health care was mentioned as a problem by a minority of respondents. Only about 10 percent of the respondents in the TLSS 2007 mentioned that they had not sought care when they needed it 74 because care was too far away. What the 2007 data also show is that this factor is more of a rural concern than an urban concern because, while only 1.3 percent of urban-based respondents mentioned this as a reason, about 13 percent of rural residents mentioned it. The difference between rural poor and rural non-poor residents was very small. Access to safe water remains a key issue. Thirty-eight percent of respondents mentioned that their main source of drinking water for the household was water from a lake, stream, a river, or an unprotected dug well. This is indeed an issue of concern, especially for those rural residents for whom these are the main sources of water, irrespective of whether they are poor or not (Table 7.7). These sources of water are usually contaminated, leading to a high risk of gastric illnesses and other waterborne diseases including cholera and typhoid. Table 7.7: Main Source of Drinking Water for Dwellings Income Unprotect- Public tap/ Hand pipe Protected, standpipe dug dwell plumbing plumbing or spring dwell or Quintiles Tanker ed, dug stream Urban (local) Rural Lake, river, truck i Quintile 1 23.8 18.7 12.6 4.9 4.6 3.1 2.2 29.8 2 18.3 17.7 9.2 8.6 2.7 2.9 1.6 38.6 3 16.9 16.9 6.8 10.2 3.9 1.4 1.1 41.7 4 22.6 17.9 3.9 6.0 4.3 1.9 1.6 40.8 5 32.2 16.9 4.8 6.6 3.2 1.8 2.7 30.4 Urban poor 80.8 0.00 4.54 4.4 0.11 0.24 2.6 7.34 Urban non-poor 86.3 0.00 3.03 1.81 0.19 0.00 2.2 5.9 Rural poor 0.00 23.5 12.1 8.5 5.0 3.5 1.4 45.2 Rural non-poor 0.00 24.9 5.5 9.1 5.2 2.4 2.0 49.4 Total 22.6 17.7 7.6 7.3 3.7 2.2 1.9 36.2 Source: TLSS 2007. Access to adequate hygiene facilities is another major challenge. The data showed that about a third of the population does not have access to reasonable hygiene facilities (Table 7.8). The majority of those who do not have access to hygiene facilities were rural residents, and were almost equally distributed among rural poor and non-poor. 74 The respective number was 3 percent in TLSS 2003. 95 Table 7.8: Distribution of Respondents by whether they Had Health and Hygiene Facilities or Not Income Quintiles Health and Hygiene Facilities Yes No Quintile 1 57.8 42.2 2 68.3 31.7 3 66.6 33.4 4 66.7 33.3 5 69.9 30.1 Urban poor 78.3 21.7 Urban non-poor 75.4 24.60 Rural poor 59.7 40.3 Rural non-poor 63.9 36.2 Total 65.8 34.3 Source: TLSS 2007. Food security is an ongoing concern. The most recent World Food Programme/Food and Agriculture Organization (WFP/FAO 2008) analyses showed that up to one-third of the rural population face food insecurity--12 percent severely (600,000 people) and 22 percent moderately (1.1 million people). In urban areas, 200,000 people are severely food insecure (15 percent) and 300,000 are moderately insecure (22 percent). The impact of food shortage is clearly seen in the incidence of malnutrition among children. For many years, the incidence of acute malnutrition among children has been high, and though levels of chronic malnutrition have decreased over the last decade, they remain high, at 27 percent (UNICEF 2007). Additional analysis undertaken by UNICEF shows that 17 percent of children under age 5 in Tajikistan are underweight (low weight for age); 7 percent of young Tajik children are wasted (low weight for height), and 27 percent of children under age 5 are stunted (low height for age). In addition, an estimated 35 percent of the population is iodine deficient, a deficiency that results in disorders such as goiter and intellectual retardation. The prevalence of anemia caused by iron deficiency is also an issue of concern, especially for pregnant and lactating mothers. Conclusions Health indicators in Tajikistan continue to be among the lowest in the Europe and Central Asia Region. The poor continue to be much more disadvantaged in terms of access to health care, and they have worse health outcomes compared to the non-poor. It is clear that poverty in general is an important determinant of health outcomes, as are other factors such as education and availability of clean water. This indicates that the Ministry of Health (MOH) needs to have the tools to collaborate with sectors that are in charge of improving nutrition, education, transport, clean water, and sanitation facilities. Over the past several years, there has been some progress toward both achieving some MDG targets and providing improved access to health services. At the national level, trends in infant and child mortality are heading in the right direction--downward. However, differences by socioeconomic characteristics show that the poor continue to be much more disadvantaged. A lot still needs to be done to ensure that progress toward meeting MDG targets is maintained. It is clear that various management, financing, and expenditure weaknesses continue to affect the health care system, and these deficiencies more greatly impact the poor. Low public financing, excessive and outdated hospital infrastructure, underfunded primary health care, and inefficiency in resource allocation all contribute to poor health outcomes. Without reforms to improve the efficiency with 96 which meager resources are used, health outcomes will continue to be poor. The Government began reforming the health sector as early as the 1990s, and the MOH continues to work closely with development partners, for example, in implementing a Health Financing Strategy, which aims to tackle financing, allocation, and ways that health care providers are paid. In the last few years, the Government has started to consistently pursue some of the recommended reforms needed in the health sector. The Government has introduced changes that aim to decrease the proportion of population for whom accessibility of health care is prohibited by cost, to decrease the share of health costs in household expenditure, and to improve equity in distribution of the health budget by introducing per capita financing for primary health care and case-based payments for hospitals. However, there is evidence that despite the Government's increased allocations to the health sector, these allocations are not sufficient to meet the needs of the sector. The TLSS 2007 shows that the bulk of health care expenditure is still borne by households. The cost of health care continues to be a deterrent to accessing care, especially by the poor. Nevertheless, the overall tendency not to seek health services associated with the cost of treatment has decreased since 2003, due to improvements in household welfare. To achieve further progress, the population needs to be better informed of the costs of care and the choices they have in terms of where to access care. Such practices are being piloted in the Basic Benefits Package (BBP) pilot rayons. The review of the quality of data on outcomes has shown that the health information system needs to be improved. For example, while non-communicable diseases constitute the largest burden of ill health, there is hardly any data on this area. In addition, data on maternal and child mortality vary significantly from source to source either because of poor quality of vital registration or due to different survey methodologies, all which add to the lack of clarity on key outcome measures. Recent or planned data collection efforts should improve the situation regarding health statistics. Those efforts include the TLSS 2007, which provides a good baseline for measuring the impact of the BBP on some intermediate outcomes. The Community and Basic Health Project is supporting the collection of baseline data and follow-up surveys to ascertain the impact of the introduction of the BBP. In addition, UNICEF has commissioned specific surveys to ascertain the impact of reforms on women's and children's health. These and other sources should provide ample opportunity to review progress in how reforms mentioned in this chapter are impacting the poor. 97 Annex 1: Description of the 2007 Tajikistan Living Standards Survey There were two previous Tajikistan Living Standards Surveys (TLSS) done in Tajikistan, one in 1999 and another in 2003. In 2007, the World Bank and United Nations Children's Fund (UNICEF) collaborated on a third TLSS, with the fieldwork implemented by the National Committee for Statistics (GosKomStat). The questionnaire design was based on the LSMS survey that was fielded in 2003, with additional questions and modules. The same households were not visited in all surveys. In addition, the sample designs used in the three years were different, although all three surveys are representative at the national, urban/rural, and regional levels. Fieldwork for the TLSS 2007 was done in two stages. The first stage was done in September­October 2007 during Ramadan. The second stage was done during October­November 2007. The sample was selected using a stratified two-stage cluster design. The total number clusters of the TLSS 2007 was 270 and the total number of households in the sample was 4,860 (270 clusters, 18 households, in each). Of the 4,860 households in the First Round, 4,490 households were revisited in the Second Round. The First Round questionnaire collected complete information from the household on various topics. The Second Round questionnaire collected information on changes to the roster since the First Round, migration information for those members who joined the household after the First Round, anthropometrics75 for children less than 6 years of age, and additional health expenditures. It also readministered the food consumption and food security modules. Three questionnaires were used to collect information for the TLSS 2007: a household questionnaire, a female questionnaire for recording information about women of child-bearing age, and a community questionnaire. These questionnaires were based on the TLSS questionnaires used in 2003, with some changes. Questions were added to existing modules and new modules were added on Migration, Financial Services, Subjective Poverty and Food Security, and Subjective Beliefs. The TLSS 2007 has also administered a module on HIV/AIDS awareness and conducted Immunizations and Anthropometric Measurements for children aged 0­5. The Labor Market Module was changed substantially from 2003, to better explore the informal labor market. The food expenditures module included additional food products. The detailed contents of the survey instrument are provided in the TLSS 2007 Basic Information Document, which is publicly available. 75 Anthropometrics is the measurement of humans. 98 Annex 2: Calculation of Consumption Aggregate The purpose of constructing a consumption aggregate is to assess the economic well-being of the population. As an index of welfare, consumption is generally preferred to income since the former provides a more adequate picture of actual well-being, especially in low- or middle-income households. The consumption aggregate as created from the 2007 Tajikistan Living Standards Survey (TLSS 2007) combines actual consumption of food with expenditures on non-food. The Components of the Consumption Aggregate In order to be a good welfare predictor, the consumption aggregate must be as comprehensive as possible, and the TLSS 2007 collected the necessary information to calculate all the main components of the consumption aggregate: food consumption (both purchased and non-purchased), non-food expenses (clothing, household articles, and so forth), utilities (water, gas, telephone, electricity, and so forth), education, health, durables, agricultural production, and housing. Although food and non-food expenses were recorded for different reference periods, all expenses were adjusted to be expressed in monthly terms. For a hypothetical food or non-food item "X," for example, its monthly equivalent M_X is computed as follows: M_X=X if the reporting time unit is month/last month M_X=X*30 if the reporting time unit is day M_X=X*4.33 if the reporting time unit is week/last seven days M_X=X / 3 if the reporting time unit is a quarter M_X=X / 6 if the reporting time unit is semiannual/last seven months M_X=X / 12 if the reporting time unit is year/last year. Food Consumption Food consumption information was collected by the food consumption module. The module contained questions on both purchased and non-purchased portions of consumption. The non-purchased portion was further subdivided into four subsections: 1. Own produced 2. Received as gift /humanitarian donation 3. Received as part of wages 4. Taken from stock. The reference period was "last seven days." Since households often buy items in bulk, and the item purchased in the last seven days may not have been consumed in the last seven days, another question was asked regarding how much of the amount purchased was consumed.76 This information was used to correct for bulk expenses and adjust the food expenditure in the same reference period. 76 The questions regarding food consumption were changed between the First Round and the Second Round of data collection. See Appendix 4 for more information. 99 There were 59 food items included in the TLSS 2007. Meals and alcoholic and non-alcoholic drinks consumed outside the home were also included in the food module. The total number of transactions recorded for the 4,860 households in the First Round was 81,974. The total number of transactions recorded in the Second Round was 83,534. Non-food Expenses Miscellaneous Expenses Households were asked to recall expenditures on a number of non-food items in Module 11 of the main questionnaire. These included, among others, clothing, household supplies for cleaning, household articles, books, entertainment, services. Since these expenses generally occur with different frequencies, to ensure coverage, the questionnaire asked the households to recall their expenditure on these items using three different periods of reference: · The last 30 days, · The last six months, and · The last year. However, as noted, when included in the consumption aggregate, all these expenses were adjusted appropriately to be expressed in monthly terms. Gambling expenses and the cost of ceremonies were left out of consumption aggregate. Education Education expenses were collected at the individual level and at the household level for expenses on part- time courses for adults and children. Expenditures for education included all education-related expenses from preschool to higher education: transportation, school fees, uniforms, textbooks, meals and lodging, in-kind or cash contributions to school, and private tutoring. Education expenses were recorded using two different periods of recall: the school/academic year and the last month. Since all other educational expenses were reported for the previous academic year, they were simply divided by 12. Health The questionnaire included an extensive health module (Module 4), and health expenditures were recorded for consultation, medicines, laboratory exams, hospitalization charges, gifts to medical personnel, transport, and other costs related to health. The health module in the First Round was supplemented by the health module in the Second Round, the latter containing questions on expenditure on prescribed and non-prescribed medicine.77 A choice was made to use the Second Round on questions related to utilization of outpatient health care. The survey included questions on not only individual components but also the total expenses as a check on the numbers. In a small number of cases where the total exceeded the sum of the individual 77 As detailed in Appendix F of the Basic Information Document (BID) associated with the TLSS 2007, outpatient health expenditures (Module 4b) were recollected because expenditures on medicines had been inadvertently left out of the First Round Main Questionnaire. 100 components of expenditure, the higher figure was used. Individual components based on the average shares were also imputed to fill in the missing values. Hospitalization charges were excluded from the consumption aggregate. The question about hospitalization expenses related to the "last" episode of hospitalization over the last 12 months and did not lend itself to be interpreted in terms of a specific period. Utilities Extensive information on expenses for utilities was collected in Module 7, which also included detailed questions about the household's dwelling. Utilities in the questionnaire included electricity, gas, telephone services (landline telephone, mobile, and other--such as public--phone), water and sewerage, fuels (firewood, kerosene, diesel, and so forth). Information on water bills was reported for various frequencies, ranging from weekly to yearly, and was adjusted appropriately to convert them into monthly expenses. In addition to electricity, other sources used for heating and lighting included gas, firewood, coal, oil/kerosene/diesel fuel, dung, and cotton stocks. The questionnaire asked for a typical monthly expenditure in the summer period and during winter. The average monthly expenditure was calculated taking the mean of the summer and winter typical month (since it was assumed that winter and summer have the same length). In a small number of cases, the household was unable to give the value of expenditure, despite reporting consumption of that item. In those cases, values were imputed with the median expenditure of households living in the same rayon/oblast and making use of the same alternative sources for heating and lighting. Transfers Charitable contributions and donations (both cash and in-kind) to other households were also included in the consumption aggregate. Housing The value of the rent of a dwelling is an index of the benefit that the household obtains from living in a certain dwelling. However, it is particularly difficult to express this benefit in monetary terms. In Tajikistan, the percentage of households that actually rent the dwelling where they live is likely to be very small. To overcome this obstacle, the survey asked the household head to report a likely rental value if the household had rented their house to others. Neither actual nor estimated rents were included in the aggregate because the inclusion of implicit rents would significantly overestimate the consumption of those renting in their dwelling, compared to households owned by members. Durable Goods Expenditures on durable goods are excluded from the consumption aggregate because the value of the purchase would distort household rankings and the poverty profile. Expenditures on semi-durable goods and small appliances are included in the consumption aggregate. 101 Agricultural Production Some agricultural expenses were included in the consumption aggregate. These include payment for the use of land, expenditure on seeds and other costs, and expenditure incurred on livestock. These were divided by 12 to arrive at the monthly value. Adjusting for Regional Price Differences Nominal expenditures are affected by substantial price differences between urban and rural areas, and between different parts of the country. Thus, such differences need to be corrected. This adjustment was made using information collected in the main household survey (using the budget share collected in the survey and the implicit prices or unit values of food items). A separate price index was also calculated using prices collected in the community questionnaire. Adjustment by Stratum-level Regional Deflators The nominal consumption aggregate is not used for the main welfare measure. It was first deflated by the stratum-level regional price deflator, which was developed based on food survey prices and the stratum- level differences of the food basket compared to the country average prices. The State Statistical Committee of Tajikistan (GosKomStat) calculates the Consumer Price Index (CPI) based on prices only in urban areas in different regions of Tajikistan. Taking into account that the official CPI does not reflect the price differential in rural areas, the decision was made to create price indexes on regional and urban/rural levels based on the TLSS 2007 data. The price index was created based on food prices only, because data about non-food expenditures is not sufficient to get proper price indexes for non-food items. The food price index (FPI) for each transaction (purchase or consumption from own production) was calculated as a transaction price divided by the national average price. Then the weighted average of the FPI for each stratum was calculated. The weight was the "importance" of the product, that is, the total value of the products spent by all households in the country. Figure A2.1 presents the normalized food price index for regions and urban/rural areas, when the county average FPI for the survey period is estimated as 1.000. 102 Figure A2.1: Normalized Food Price Index Note: Estimated using TLSS 2007 data. The nominal food consumption of household was deflated by the regional price deflator FPI using the following formula: Deflated food consumption = Nominal food consumption / Regional price deflator. 103 Annex 3: Construction of the Poverty Line Based on the Cost of Basic Needs (CBN) Approach Previous Tajikistan poverty assessments used international poverty lines equivalent to Purchasing Power Parity (PPP) US$2.15. The international poverty lines usually are most suitable for comparison of poverty statistics among countries. However, in addition to using international poverty lines, many countries are using national poverty lines, which more accurately reflect the cost of the objectively defined cost of minimum needs. Tajikistan has not officially adopted a minimum consumption basket, and this is why the decision was made to derive such poverty line from the survey data. To develop the poverty line for the analyses of the Tajikistan Living Standards Survey (TLSS) 2007 data, we use the Cost of Basic Needs (CBN) approach. The following basic steps are followed: · Identifying a reference group from which food consumption patterns can be drawn. A fixed nominal expenditure level is used to define the reference group. Given that food costs appear to be similar across regions, setting one expenditure range to define the reference group will establish a similar living standard with regards to food across regions. · Setting the calorie requirements. Recommended calorie needs are derived using the World Health Organization (WHO) caloric requirements. · Deriving the food poverty line by calculating the caloric value unit, which is the cost of each calorie the reference group consumes. · Estimating the allowance for non-food goods. The non-food allowance is anchored to the consumption behavior of the poor within each sector. We employ two standards for calculating the non-food portion of the poverty line. For the complete poverty line, we estimate the nonfood amount based on households whose food expenditure is just equal to or a little more than the food poverty line. Each of those steps is described in greater detail below. (i) The choice of reference population The choice of reference population for the food poverty line is guided by the expectation that it will reflect the population of households near the poverty line--thus reflecting food consumption that is near the poverty line (reflecting a minimum food basket that is not "too" poor and not overly rich). The choice of the reference population is a normative judgment in the construction of a poverty line. Ideally, the reference group will be chosen to be consistent with the resulting poverty estimates based on behavioral parameters of the reference group. In theory, then, one must first approximate who are the poor to set the reference group and then calculate the poverty line. In some cases, it is necessary to iterate until there is convergence, by revising the reference group accordingly. In this analysis, the reference population to set the food consumption pattern is the population of people in the 4th and 5th deciles of the per capita consumption distribution among all individuals. The food basket of this group is meant to capture the food consumption patterns for a relevant, relatively low- income population. 104 (ii) Calorie requirements, composition of minimal food basket, and calorie cost We calculate the food poverty line as the cost of buying a diet of 2,250 calories per capita per day, given the food consumption patterns of households in a reference population. For each food item f, a caloric content value, cf, is assigned based on calorie tables produced by the United Sates Department of Agriculture (USDA). There are 59 food and beverage items in the TLSS 2007 food consumption module (not including alcoholic beverages and food eaten outside). For each food item, the share of total calorie intake, Sf, is computed. Based on the consumption shares of this reference population, 2,250 calories per day is then allocated across the most important food items constituting 98.5 percent of the total food consumption basket for this group. This resulted in using 36 food items out of 59 recorded in the food module of food consumption of TLSS. The decision to not include all the 59 food items in the minimum food basket is driven by the consideration that the food items that have a small share in the basket and low frequency of purchase have greater sampling error and lower reliability. This minimum calorie diet is then priced by mean national prices using the price per calorie (Pf/cf) for each food item. Sensitivity analysis proves that in the same reference population the cost of the basket is robust to using a basket consisting of a different number of the most important products. For instance, the cost of the 2,250 calories for the same reference population of the 4th and 5th population deciles costs just 0.02 somoni per day more if we use all 59 products instead of 36 products. (iii) The food poverty line The food poverty line is then computed as the total cost of this reference population minimal food basket. The food poverty line (FPL) can simply be expressed as: Pf f FPL = S (2,250) . f cf The prices for each food (Pf) are drawn from the national unit value prices calculated from the food diary. Using this methodology, the food poverty line is estimated as 2.92 somoni per capita per day needed to obtain 2,250 calories per day.78 The monthly value of the food poverty line is equal to 88.8 somoni per capita. The composition of the cost of the minimal food basket is provided in Figure A3.1. 78 The choice of the reference population in the lower part of the distribution results in a similar value of the food poverty line. This proves that the poverty line calculated based on the reference population in the 4th and 5th deciles leads to robust estimates of poverty. 105 Figure A3.1: The Composition of the Cost of a Minimal Food Basket Composition of the the cost of Minimal Food Basket Coffee, tea and Spicies and other cocoa Mineral water, soft Sugar, jam, 0.6% 1.5% drinks, juices honey, chocolate 0.2% and confectionary 6.7% Vegetables Bread and cereals 11.3% Fruit 43.6% 7.6% Meat Butter, Oils and 9.8% fats Milk cheese and Fish 12.4% eggs 0.1% 6.3% Source: World Bank estimates using TLSS 2007 data. (iv) Total (complete) poverty line The need for non-food consumption requires adding an allowance for non-food goods and services to the food poverty line. The upper-bound method used here to determine the value of the general or complete poverty line (CPL) was developed by M. Ravallion (see Ravallion 1994). To determine the allowance for non-food consumption, using the data itself, first those households whose per capita food consumption is just above the value of the food poverty line (that is, when per capita consumption is within the interval [Food line, Food line + 20 percent]) are selected. This part of the sample will constitute the reference group for the derivation of the general poverty line. The share of total consumption that goes to non-food consumption is calculated for this reference group. This share is the "allowance" for non-food consumption that is added to the value of the food poverty line to get the total (complete) poverty line. We estimate that if the share of food consumption among those whose total consumption is just above the value of the food poverty line is 64 percent, then non-food consumption represents 36 percent. The value of the total (complete) poverty line is then derived as a sum of the value of the food poverty line plus the value of non-food consumption. It is equal to 4.56 somoni per day per person or 138.8 somoni per month per person.79 This monthly value consists of the food poverty line value of 88.8 somoni and the non-food component value of 50 somoni. Such method of deriving the complete poverty line is the simplest way to assess the value of the minimum consistent with the consumption patterns of the reference population. We have chosen the simplest method described above, because it is the most transparent, most easily replicable, and most intuitive, which should help with national poverty diagnostics.80 79 Daily values are transformed into monthly values using the factor of 30.4 (average number of days per month). 80 The poverty line does not contain the components of housing rents and durable goods. Thus, it is appropriate to use only for the comparison with the consumption aggregate, which also excludes components of housing rents and durable goods. 106 Annex 4: Welfare Dynamics Based on the Consumption Aggregate: Comparing 2003 and 2007 Data The survey-based welfare (consumption) comparisons between 2003 and 2007 are not straightforward for the following reasons: · Different time of the year the survey was conducted; 2003 (May­June) compared to 2007 (October­November); Tajikistan monthly data from the Household Budget Survey (HBS) suggests that autumn consumption is 15 to 20 percent higher than summer consumption. · The list of consumption items (including food) is more extensive in 2007 (also, some items that were aggregated in 2003, for example, "other vegetables," compared to specific items in 2007); the list was extended in 2007 to more fully gauge food consumption (also recommended by the Food and Agriculture Organization [FAO]). · Which prices to use for adjustments over time? Gross domestic product (GDP) deflator inflation suggests an inflation of 97.3 percent between 2003 and 2007; Consumer Price Index (CPI) inflation is 52.6 percent; costing of food items based on the 2003 and 2007 surveys suggest that food prices increased by more than 100 percent. To be able to compare the levels of poverty between 2003 and 2007, we make the following adjustments (to the local currency value of the Purchasing Power Parity [PPP] US$2.15 poverty line in 2003) dictated by the data: · A 106.9 percent increase in prices between June 2003 and 2007: (Note: the food price index derived directly from the survey indicates that food prices increased by 141 percent (while official food CPI suggest an increase of 60 percent); increase in the non-food prices was 30.2 percent, and increase in services was 97.4 percent; using weights for food, non-food, and services components of, respectively, 61 percent, 25 percent, and 14 percent, we get an overall increase in prices of 106.9 percent. (Note: GDP deflator over the same period of time suggests an increase in prices of 105.6 percent, which is very close). · A 16.4 percent increase due to various times of the year the surveys were conducted (this adjustment factor was derived empirically based on the monthly data from the 2005 panel of HBS households by comparing the value of November consumption to that of June consumption in real terms). · A 6.0 percent increase due to using a more extensive list of food consumption items in the 2007 survey (this adjustment factor was derived by comparing the "full" consumption aggregate to the "restricted" [same items as in 2003] consumption aggregate). · By combining these three adjustment factors we get an overall adjustment factor of a 155.3 percent increase (that is, the price index is 255.3 percent), which, applied to the value of the PPP US$2.15 poverty line in 2003 (47.1 somoni) gives a value of 120.1 somoni (comparable with the 2007 survey consumption aggregate!) What is the estimated reduction in poverty between 2003 and 2007 using the PPP US$2.15 poverty line? Poverty headcount declined from 63.5 percent in 2003 to 40.9 percent in 2007, or by 22.6 percentage points (this result implies the poverty-growth [per capita] elasticity of -1.3 percent); note that if we used the official CPI, the poverty decline between 2003 and 2007 would be even higher. 107 Annex 5: Sensitivity of Poverty and Poverty Profile to Using an Equivalence Scale A. Adult Equivalent Consumption ­ General Formula The welfare literature recognizes that there may be economies of scale in consumption. For example, a two-person household does not imply double expenditures on housing, utilities, or other non-food items for which consumption can be shared (these are public goods whose cost does not vary whether one person or a number of people use the good). Larger households might also be able to buy food or nonfood items in bulk, which can mean lower prices or discounts. Moreover, the age structure of household members, where a child is assumed to not be equivalent to an adult in terms of needs, could be considered. Consider the adult equivalent household size to be defined by the general formula: AE=(A+K). In this formula, captures the economies of scale in consumption. The parameter identifies the weight to convert each child into an equivalent adult. (Note: Per capita scale is the particular case of the general scale with both parameters equal to 1.) The adjustment for household size and composition is done by dividing total household consumption by AE (per-adult-equivalent household size), which gives an adjusted measure of consumption. For example, a household with an adult equivalent size of 3.5 needs to spend 3.5 times as much as a single adult in order to be equally well off as the single adult. The choice of equivalence scale reflects judgments about differences in needs. For example, the per capita adjustment incorporates the extreme judgment that all household members have equal needs irrespective of age. In the Republic of Tajikistan, food is a large share of household consumption. Since food is generally not associated with economies of scale, using a per capita adjustment that has no economies of scale seems plausible.81 On the other hand, taking into account that the food needs of children are lower compared to that of adults, the parameter alpha between 0.5 and 1 may be considered as quite logical. The consumption aggregate also includes some pseudo shared items such as expenditures on certain services, so the formula with the parameter closer to 1 like 0.8 will be plausible, too. B. Adjustment of Scales Based on the Parameters of Modal Household Structure As pointed out by Deaton and Zaidi (1999), the adjustments of household consumption using adult equivalent size would overestimate the total consumption unless all households were single-adult households. They suggest using an adjusted adult equivalent size that will keep the modal or pivotal households in per capita dimensions. Hence, the adjusted adult equivalent size of the household i (AE_ADJi) is defined as: HHSIZE mod al AE_ADJi = AEi = AEi . AESIZE mod al 81 The exception to the notion that food does not have economies of scale would be bulk-purchase discounts. If certain foods are perishable and the cost of storage is high, then large households may be better able to take advantage of bulk-purchase discounts. 108 For the modal households, the household size and the adjusted adult equivalent size are the same. For example, in the case of the general formula with the parameters = 0.7 and theta = 0.8: A0 + C0 AE_ADJ i = AEi = AEi * 1.51, ( A0 + C0 ) where A0 and C0 are the numbers of adults and children in the typical "modal" household, respectively, and Ai and Ci are the numbers of adults and children in the ith household. The modal or pivotal household is a five-member household with two adults and three children (A0 = 2 and C0 = 3). The idea of the adjustment by the parameters of a modal household structure is as follows. If the poverty line in our analysis was calculated for one person (per capita), but not for one single adult family, and if we assume the economies on scale, then this kind of adjustment keeps modal families in the per capita dimension with any adjusted scale. Otherwise, using the scale without such an adjustment will in fact apply the poverty line calculated for one person to the single-person household, which will lead to the overestimation of consumption and the underestimation of poverty rates. Thus, in this analysis only adjusted scales are used for the poverty analysis. Figure A5.1 compares the poverty statistics using per capita and per adult equivalent (adjusted with the modal household parameters). While the overall poverty rate remains almost unchanged, (unlike the case of using non-adjusted adult equivalence scales like in the case of the Organisation for Economic Co- operation and Development [OECD] scale), the poverty profile is different by household size. When we use the OECD scale, we get only a slightly lower extreme poverty headcount (16.4 percent compared to 17.1 percent with no use of scale). 109 Figure A5.1: Poverty by Household Size (per capita compared to per adult equivalent) Panel A Poverty headcount index per capita vs. adult equivalent 0.700 0.600 Poverty headcount 0.500 index per 0.400 capita 0.300 0.200 Poverty 0.100 headcount index adult 0.000 equivalent 1 2 3 4 5 6 7 8 9 10+ Household size Panel B Extreme Poverty Headcount Index per capita vs. adult equivalent 0.300 Extreme poverty 0.250 headcount 0.200 index per capita 0.150 0.100 Extreme 0.050 poverty headcount 0.000 index adult 1 2 3 4 5 6 7 8 9 10+ equivalent Household size 110 Annex 6: Sensitivity of Poverty to Pricing of the Flour/Bread Panel A: Original Poverty Estimates Per capita Monthly per consumption capita adjusted with consumption Strata Price nominal Deflators Oblast Location p0f p1f p2f p0 p1 p2 SOMONI SOMONI Dushanbe URBAN 0.164 0.038 0.013 0.433 0.132 0.056 201 183 Total 0.164 0.038 0.013 0.433 0.132 0.056 201 183 Sogd URBAN 0.245 0.064 0.025 0.536 0.191 0.087 205 198 RURAL 0.333 0.067 0.020 0.740 0.244 0.102 129 129 Total 0.311 0.066 0.021 0.688 0.231 0.098 148 147 KHatlon URBAN 0.144 0.026 0.008 0.525 0.137 0.050 151 158 RURAL 0.068 0.009 0.002 0.462 0.096 0.029 146 154 Total 0.082 0.012 0.003 0.473 0.103 0.033 147 155 RRP URBAN 0.248 0.043 0.012 0.568 0.185 0.074 167 163 RURAL 0.125 0.025 0.007 0.476 0.123 0.045 163 162 Total 0.140 0.027 0.007 0.488 0.131 0.048 163 162 GBAO URBAN 0.008 0.000 0.000 0.184 0.039 0.011 263 236 RURAL 0.112 0.020 0.007 0.472 0.119 0.043 176 154 Total 0.098 0.018 0.006 0.434 0.109 0.038 187 165 Total URBAN 0.189 0.043 0.015 0.494 0.154 0.065 188 180 RURAL 0.164 0.031 0.009 0.550 0.149 0.056 146 149 Total 0.171 0.034 0.010 0.535 0.150 0.058 157 157 Panel B: Simulated Poverty Estimates (pricing self-produced bread at the cost of flour rather than a market price of bread) Per capita Monthly per consumption capita adjusted with consumption Strata Price nominal Deflators Oblast Location p0f p1f p2f p0 p1 p2 Somoni Somoni Dushanbe URBAN 0.188 0.045 0.016 0.450 0.146 0.063 196 179 Total 0.188 0.045 0.016 0.450 0.146 0.063 196 179 Sogd URBAN 0.274 0.074 0.030 0.557 0.208 0.098 201 194 RURAL 0.416 0.096 0.031 0.782 0.286 0.130 121 121 Total 0.380 0.090 0.031 0.724 0.266 0.122 141 140 KHatlon URBAN 0.214 0.040 0.013 0.585 0.173 0.068 141 148 RURAL 0.127 0.019 0.005 0.591 0.137 0.046 136 143 Total 0.142 0.023 0.006 0.590 0.143 0.050 137 143 RRP URBAN 0.290 0.061 0.018 0.613 0.214 0.093 159 155 RURAL 0.171 0.036 0.011 0.549 0.156 0.061 152 152 Total 0.186 0.039 0.012 0.557 0.163 0.065 153 152 GBAO URBAN 0.055 0.003 0.000 0.246 0.051 0.015 251 226 RURAL 0.170 0.030 0.009 0.557 0.149 0.056 165 145 Total 0.155 0.027 0.008 0.517 0.136 0.051 177 155 Total URBAN 0.228 0.053 0.019 0.527 0.176 0.077 182 174 RURAL 0.227 0.047 0.015 0.636 0.187 0.075 137 139 Total 0.227 0.049 0.016 0.607 0.184 0.076 149 148 Note: P 0(1) (2) refer to the FGT poverty measures (poverty headcount, poverty gap, and poverty gap squared). "f" stands for the food or extreme poverty. 111 Annex 7: The 2003­07 Growth Incidence Curves Figure A7.1: The 2003­07 Growth Incidence Curves Total (years 2007 and 2003) Urban Growth-incidence 95% confidence bounds 16 16 Growth in mean Mean growth rate A n n u a l g r o w th ra te % A n n u a l g r o w th ra te % 12 12 8 8 4 4 0 0 1 10 20 30 40 50 60 70 80 90 100 1 10 20 30 40 50 60 70 80 90 100 Expenditure percentiles Expenditure percentiles Rural 16 A n n u a l g r o w th ra te % 12 8 4 0 1 10 20 30 40 50 60 70 80 90 100 Expenditure percentiles Source: World Bank estimates using TLSS 2003 and 2007 data. 112 Annex 8: Description of Methodology 2 and the Results Methodology The simulations (Methodology 2) (Table A8.1) are based on two assumptions, which are required to estimate the poverty impact of a return of migrants in the absence of a real-world counterfactual. Naturally, the simulation has to be understood as an "as-if" thought experiment, which is very unlikely to happen in reality, since many confounding parameters, which determine the level of poverty, have to be held constant, although they are very likely to change during an economic crisis. Thus, our estimates are a partial equilibrium simulation of the pure effect after the return of migrants, given a set of limiting assumptions. It is important to acknowledge that we perform an ex post analysis after return, that is, we do not consider the bringing home of any leftover savings. Also, we will not allow for job search of returnees (that is, return migrants), but simply assume that a specific share of migrants finds a paid job immediately after return. First, we have to select a specific number of migrants who will be considered returnees. In the absence of information about which migrants are more likely to return in reality--most likely the return probability will be determined by the planned length of stay, the actual time already abroad, the sector of employment, the validity of legal documents, the economic situation of the household at home, potential savings, age and "soft" factors like homesickness, and so forth--we decide to perform a random draw of one-fifth, one-third, and one-half of migrants. As these migrants return, the household does not receive any further remittances (unless another, non- returnee migrant belongs to the same household) but faces higher consumption requirements. Since most migrants from Tajikistan are prime-aged males, we account for one additional consumer in the calculation of required monthly adult equivalence household consumption. The probability of being employed is predicted from working-age population regressions on a wide range of individual and regional characteristics. We omit the option of working in farming because it normally does not generate income and because farming activities are predetermined (the land size and the seeding are assumed to be exogenously given at the time of a migrant's return). Working returnees will earn an average monthly net wage, predicted from earnings regression for the working population of Tajikistan. 113 Table A8.1: Simulation Results (Methodology 2): Poverty Headcount and Poverty Gap Poverty Headcount Poverty Gap Urban Rural Total Urban Rural Total Dushanb e Situation 2007 42.7% 42.7% 13.0% 13.0% Simulation minus 20% of migrants (2007) 44.1% 44.1% 14.2% 14.2% Simulation minus 33% of migrants (2007) 44.3% 44.3% 15.1% 15.1% Simulation minus 50% of migrants (2007) 45.1% 45.1% 16.0% 16.0% Sogd Situation 2007 53.6% 74.3% 69.0% 19.2% 25.0% 23.5% Simulation minus 20% of migrants (2007) 54.7% 75.8% 70.4% 21.0% 27.8% 26.1% Simulation minus 33% of migrants (2007) 55.4% 76.8% 71.3% 21.3% 30.2% 28.0% Simulation minus 50% of migrants (2007) 56.0% 77.0% 71.6% 22.0% 32.0% 29.4% Khatlon Situation 2007 52.5% 45.8% 47.0% 13.7% 9.6% 10.3% Simulation minus 20% of migrants (2007) 52.7% 47.6% 48.5% 13.8% 12.4% 12.6% Simulation minus 33% of migrants (2007) 53.6% 47.8% 48.8% 14.9% 12.8% 13.2% Simulation minus 50% of migrants (2007) 54.9% 49.3% 50.2% 15.5% 14.5% 14.7% RRS Situation 2007 56.8% 46.1% 47.4% 18.5% 11.8% 12.6% Simulation minus 20% of migrants (2007) 59.4% 48.4% 49.7% 23.2% 15.6% 16.6% Simulation minus 33% of migrants (2007) 60.0% 50.1% 51.4% 23.0% 17.7% 18.4% Simulation minus 50% of migrants (2007) 59.5% 50.7% 51.8% 23.8% 19.6% 20.1% GBAO Situation 2007 19.3% 46.5% 42.9% 4.0% 12.1% 11.1% Simulation minus 20% of migrants (2007) 23.7% 49.9% 46.5% 8.8% 17.9% 16.7% Simulation minus 33% of migrants (2007) 25.5% 54.1% 50.4% 10.7% 21.8% 20.3% Simulation minus 50% of migrants (2007) 31.7% 56.0% 52.8% 16.2% 25.9% 24.6% Total Situation 2007 49.3% 54.4% 53.1% 15.4% 14.9% 15.0% Simulation minus 20% of migrants (2007) 50.4% 56.3% 54.8% 16.9% 18.1% 17.8% Simulation minus 33% of migrants (2007) 51.1% 57.3% 55.7% 17.6% 19.7% 19.1% Simulation minus 50% of migrants (2007) 51.9% 58.2% 56.5% 18.4% 21.5% 20.7% Source: World Bank estimates based on TLSS 2007. 114 Annex 9: The Regression Analysis of the Poverty Correlates Table A9.1: The Coefficients of Comparative Regression Analysis (2003 compared to 2007) 2003 2007 Urban Rural Urban Rural coef se coef se coef se coef se Household characteristics Log of household size -0.298*** 0.08 -0.497*** 0.08 -0.439*** 0.08 -0.672*** 0.07 Log of household size squared -0.061** 0.03 0.011 0.02 -0.036 0.03 0.077*** 0.02 oblast (region) Dushanbe (dropped) (dropped) (dropped) (dropped) Sogd -0.164*** 0.04 -0.312*** 0.03 0.006 0.04 -0.149*** 0.02 Khatlon -0.330*** 0.05 -0.420*** 0.03 0.153*** 0.04 0.127*** 0.02 RRP 0.037 0.06 (dropped) -0.016 0.05 0.161*** 0.02 GBAO -0.223*** 0.06 -0.651*** 0.04 0.323*** 0.06 (dropped) access to land (own/rented), used for farming (sotka) no (dropped) (dropped) (dropped) (dropped) 1 - 10 0.154*** 0.04 0.118*** 0.04 -0.010 0.04 0.119*** 0.02 11 - 20 0.149 0.16 0.143*** 0.04 0.066 0.13 0.159*** 0.03 21+ 0.392*** 0.11 0.245*** 0.04 -0.039 0.08 0.192*** 0.03 Characteristics of household head Log of household head's age -0.068 0.06 0.115*** 0.04 -0.025 0.05 0.139*** 0.03 Gender male (dropped) (dropped) (dropped) (dropped) female -0.016 0.04 -0.020 0.03 -0.028 0.03 -0.034 0.02 education of hh head None (dropped) (dropped) (dropped) (dropped) Primary 0.185** 0.08 0.152*** 0.05 0.224** 0.10 -0.010 0.04 Basic -0.004 0.07 0.089** 0.05 0.202** 0.09 0.014 0.04 Secondary geneal 0.145** 0.06 0.129*** 0.04 0.191** 0.09 0.078* 0.04 Secondary special 0.250*** 0.07 0.244*** 0.05 0.241*** 0.09 0.143*** 0.04 Secondary technical 0.316*** 0.07 0.281*** 0.05 0.275*** 0.09 0.151*** 0.04 Higher 0.405*** 0.06 0.349*** 0.05 0.408*** 0.09 0.233*** 0.04 employment of hh head not employed (dropped) (dropped) (dropped) (dropped) employed 0.097*** 0.03 0.000 0.03 0.152*** 0.03 0.038** 0.02 _cons 5.452*** 0.24 5.077*** 0.17 5.581*** 0.23 5.071*** 0.16 Number of observations 1,520 2,639 1,710 3,150 Adjusted R2 0.265 0.229 0.288 0.212 Source: World Bank calculations based on LSMS 2003 and TLSS 2007 data. 115 Annex 10: Selected Labor Market Indicators Table A10.1: TLSS 2007 Main Labor Market Variables and Definitions Variable Definition Working-age population Population of Tajikistan aged 15­64. Worked during the last 14 days or had a job but was temporarily absent from work for some Employed reason. Have already found a job that will start later. Did not work during the last 14 days and--actively looked for jobs in the past 30 days or did Unemployed not actively look. Looked for jobs in the past 30 days because believe no chance of getting jobs or that there are no jobs or waiting for busy season to begin. Long-term unemployed Those unemployed who have looked for jobs for more than 12 months. Active Employed + Unemployed. Inactive Working-age population ­ active. One who migrated to another country for at least a month since January 2004, and returned Return migrant to Tajikistan between September 2006 and October 2007. Emigrant One who migrated abroad in the past and is currently residing abroad. Small firm Firm that employs 5 workers or less. Medium firm Firm that employs between 6 and 50 workers. Large firm Firm that employs more than 50 workers. Secondary job Respondent works in more than 1 job. Definitions of Formality Affiliated to Social Security Respondent is affiliated to social security through the primary job. Permanent Respondent's primary job is permanent. Contract Respondent has signed a contract/written agreement with the employer for the primary job. Labor book Respondent's primary job is recorded in the Labor book. Source: World Bank estimates using 2007 Tajikistan Living Standards Survey data. 116 Table A10.2: Main Labor Market Indicators (in levels), 2007 Working Labor Population Force Employed Unemployed All 4,215,165 2,171,008 1,965,231 205,777 Gender Male 1,974,244 1,372,366 1,217,068 155,298 Female 2,240,921 798,642 748,163 50,479 Age 15­24 1,622,471 561,497 463,213 98,284 25­54 2,288,709 1,469,139 1,367,196 101,943 55­64 303,985 140,372 134,822 5,550 Educationa Basic or less 1,211,852 403,464 358,184 45,280 Secondary General 2,112,187 1,110,072 992,353 117,719 Technical and Vocational 451,731 337,454 312,096 25,358 Higher and Graduate 365,889 300,579 284,501 16,078 Location Rural 3,081,766 1,630,244 1,479,486 150,758 Urban 1,133,399 540,764 485,745 55,019 Region Dushanbe 399,527 182,914 165,926 16,988 Sogd 1,238,532 668,507 617,748 50,759 Khatlon 1,488,712 805,433 729,581 75,852 RRS 945,485 437,856 399,488 38,368 Gbao 142,909 76,298 52,488 23,810 Ethnicity Tajik 3,214,097 1,598,140 1,447,488 150,652 Uzbek 942,780 540,785 489,709 51,076 Russian 27,415 13,842 12,627 1,215 Kyrgyz 15,175 9,755 7,097 2,658 Tatar 3,807 2,094 2,033 61 Turkmen 596 298 298 0 Other 11,295 6,094 5,979 115 Poverty Nonpoor 2,108,266 1,076,018 985,792 90,226 Poor 2,106,899 1,094,990 979,439 115,551 Male by Age Group Male15­24 749,170 319,758 251,009 68,749 Male25­54 1,079,298 947,966 865,771 82,195 Male55­64 145,776 104,642 100,288 4,354 Female by Age Group Female15­24 873,301 241,739 212,204 29,535 Female25­54 1,209,411 521,173 501,425 19,748 Female55­64 158,209 35,730 34,534 1,196 Male by Location Male Rural 1,452,363 1,026,730 911,726 115,004 Male Urban 521,881 345,636 305,342 40,294 Female by Location Female Rural 1,629,403 603,514 567,760 35,754 Female Urban 611,518 195,128 180,403 14,725 a. Numbers may not add up to totals for education because of missing values. Source: World Bank estimates using 2007 Tajikistan Living Standards Survey data. 117 Annex 11: Description of the Migration Data This annex uses regionally representative data from the Tajikistan Living Standard Survey (TLSS), collected during October and early November 2007. The overall sample consists of 4,860 households and 32,222 individuals, of which 20,000 are in the extended working age (age 15­65) population. We chose this category of working age throughout the paper, since legal pension ages are rather low in Tajikistan and a substantial share of pensioners keeps on working to sustain their livelihood. Migration Flows One problem for investigating current migration patterns at the individual level is that TLSS 2007 does not provide any individual weights for migrants. More seriously, the retrospective migration history only covers each migrant's last migration spell. Consequently, we cannot derive correct migration flow numbers for past years. For example, the migration number for 2004 is likely to be underestimated since those migrants who moved again in 2005, 2006, or 2007 are not included in the flow number of 2004, but appear only in the very last spell. Given this serious bias in retrospective migration flows we only provided information from 2004 to 2007. Regarding internal migration, we are unable to disentangle whether internal migration took place from a rural or urban origin settlement. The rayon coding of TLSS 2007 uses the same code for some urban and rural settlements, for example, Spitamen rayon in Gorno-Badakhshan Autonomous Oblast (GBAO) has urban and rural parts, but only a single code. Migration Stocks Since we have no individual weights for currently absent household members, we are unable to calculate the "real" stock of current Tajik population residing abroad, but rather compute an unweighted share in the population and infer from the share and the total population number a proxy of total numbers of migrants. Remittances Aggregate The remittances aggregate for each household consists of three components: (a) remittances received from household members who currently live abroad (reported on a yearly basis in Tajik somoni), (b) remittances received from other persons abroad (reported on a yearly basis in Tajik somoni, excluding transfers from international organizations), and (c) remittances from past migration experience of current household members (reported on a monthly basis in U.S. dollars). The latter component consists of the product of months abroad (out of the past 12 months) multiplied by 70 percent of the monthly net earnings, which reflects the share of income generated abroad and supposed to be sent or brought home. This component is then recalculated in Tajik somoni at local official exchange rates. A Consumer Price Index (CPI) correction across months was not feasible since foreign currencies are normally stored as savings and are exchanged at times unobservable to us. As the first component (yearly remittances received in Tajik somoni) proved to be rather noisy--that is, some households reported remittances in U.S. dollars, some in somoni, and for some the remittances substantially exceeded their earnings--we chose an approximate approach, estimating the first component as a fixed share (70 percent) of U.S.- dollar income that was generated during the last 12 months. This component was then recalculated to somoni according the approach described above. For the inequality measures, we ran both estimates, omitting the households with zero consumption in the simulations and correcting zero-consumption 118 households by the per capita value of 5 or 25, respectively. The latter was performed in order to include very poor households and shows slightly higher inequality, albeit the absolute difference is very small. Simulation Methodology Returnee simulations are based on two assumptions, which are required to estimate the poverty impact of a return of migrants in the absence of a real-world counterfactual. Naturally, the simulation has to be understood as an "as-if" thought experiment, which is very unlikely to happen in reality, since many confounding parameters, which determine the level of poverty, have to be held constant, although they are very likely to change during an economic crisis. Thus, our estimates are partial equilibrium simulations of the pure effect after some migrants return, given a set of limiting assumptions. Note that we perform an ex post analysis after return, that is, we do not consider the bringing home of any leftover savings. Also, we will not allow for job search of returnees (that is, return migrants), but simply assume that a specific share of migrants finds a paid job immediately after return. First, we have to select a specific number of migrants who will be considered returnees. In the absence of information about which migrants are more likely to return in reality--most likely the return probability will be determined by the planned length of stay, the actual time already abroad, the sector of employment, the validity of legal documents, the economic situation of the household at home, potential savings, age and "soft" factors like homesickness, and so forth--we decided to perform a random draw of one-fifth, one-third, and one-half of migrants. As these migrants return, the household receives no further remittances (unless another, non-returnee migrant belongs to the same household) but faces higher consumption requirements. Since most migrants from Tajikistan are prime-aged males, we account for one additional consumer in the calculation of required monthly adult equivalence household consumption. The probability of being employed is predicted from working-age population regressions on a wide range of individual and regional characteristics (Table A11.1). We omit the option of working in farming since it normally does not generate income and because farming activities are predetermined (the land size and the seeding are assumed to be exogenously given at the time of a migrant's return). Working returnees will earn an average monthly net wage, predicted from earnings regression for the working population in Tajikistan. Table A11.1: Predicted Probabilities and Fitted Values for the Simulation Mean Median Minimum Maximum Predicted Employment Probability 34.5% 31.8% 0.1% 83.4% Fitted Net Wage (somoni, monthly) 326.1 357.0 10.5 472.5 Source: TLSS 2007; authors' calculation. 119 Annex 12: Targeting the Poor in Tajikistan: Geographic or Proxy Means Testing?82 This annex analyzes and presents a comparison of geographic and means-testing targeting in Tajikistan. The analysis is based on the 2007 Tajikistan Living Standards Survey (TLSS 2007). The rationale for having a proxy-means-testing (PMT) model is the belief that such a model might do a better job at targeting public subsidies to the poor than a much simpler scheme like geographic targeting. Our simulations indicate that in the case of Tajikistan this is not necessarily the case. The geographic simple scheme has both lower errors of inclusion and exclusion than the PMT, and also has a higher impact on poverty (measured by the squared poverty gap) than the PMT scheme. According to the TLSS 2007, the total population is 7,063,800, with a headcount ratio of 53.5 percent, an average distance to the poverty line of 15.0 percent of the value of this line, and an average squared distance of 5.8 percent of the line (see Table A12.1). Geographic Optimal Targeting The simulations presented in this paper work with a fixed budget of 5 million somoni and 10 million somoni, corresponding to 20 and 40 percent, respectively, of total payments to old-age and disability pensioners. The first scheme minimizes the squared poverty gap, assuming that the only information available is the poverty measures for each oblast, and the income for all households, but that it is not possible to identify the poor households. Then, following Kanbur (1987), the oblasts have to be ranked descendingly according to the poverty gap, and every person in the poorest oblast is given a transfer of identical size, computed to take this poorest oblast to the poverty level (gap) of the next-poorest oblast. If there is money left, every person in the two poorest oblasts (now with identical poverty gaps) receives the amount needed for these two oblasts to reach the poverty gap of the third-poorest oblast, and so on. Since this scheme is based on the minimization of the squared poverty gap, we will base the description of the results on that measure (which is sensitive to extreme poverty). Results for the headcount ratio and poverty gap can be found in Table A12.1. For the case of 5 million somoni, the squared poverty gap decreases from 5.80 percent to 5.57 percent. The percentage of poor that do not receive a transfer from this scheme (error of exclusion, see Table A12.2) is 62 percent, while 31 percent of the beneficiaries are non-poor (error of inclusion). If 10 million somoni are distributed instead, the squared poverty gap decreases to 5.35 percent. 82 This annex was prepared by Juan Carlos Parra, Consultant, HDNDE; and Quentin Wodon, Adviser, HDNDE. 120 Table A12.1: Poverty Impact of Targeting Programs in Tajikistan Panel A: All Predicted Poor by PMT Are Eligible Scheme 5 Million Somoni 10 Million Somoni Squared Squared Headcount Poverty Poverty Headcount Poverty Poverty Ratio Gap Gap Ratio Gap Gap No transfer 53.50 15.00 5.80 53.50 15.00 5.80 Optimal 53.21 14.65 5.57 52.88 14.31 5.35 Naïve 53.16 14.71 5.61 52.73 14.41 5.43 PMT 53.07 14.74 5.66 52.28 14.49 5.53 PMT + Optimal 53.06 14.68 5.59 52.57 14.37 5.41 PMT + Naïve 53.09 14.72 5.63 52.60 14.45 5.46 Panel B: Only Pensioners Predicted Poor by PMT are Eligible Scheme 5 Million Somoni 10 Million Somoni Squared Squared Headcount Poverty Poverty Headcount Poverty Poverty Ratio Gap Gap Ratio Gap Gap No transfer 53.50 15.00 5.80 53.50 15.00 5.80 Optimal 53.21 14.65 5.57 52.88 14.31 5.35 Naïve 53.16 14.71 5.61 52.73 14.41 5.43 PMT 52.92 14.71 5.64 52.36 14.44 5.49 PMT + Optimal 52.95 14.65 5.58 52.76 14.33 5.40 PMT + Naïve 52.95 14.68 5.60 52.50 14.38 5.43 Panel C: Balanced Budget ­ Only Pensioners Predicted Poor by PMT are Eligible Scheme 5 Million Somoni (20%) 10 Million Somoni (40%) Squared Squared Headcount Poverty Poverty Headcount Poverty Poverty Ratio Gap Gap Ratio Gap Gap No transfer 53.50 15.00 5.80 53.50 15.00 5.80 Optimal 52.90 14.40 5.42 53.36 14.39 5.40 Naïve 52.72 14.42 5.45 53.05 14.42 5.44 PMT 52.54 14.41 5.47 52.43 14.41 5.48 PMT + Optimal 52.73 14.42 5.45 53.24 14.45 5.47 PMT + Naïve 52.65 14.41 5.45 52.81 14.42 5.44 Note: Poverty measures in percentages. Geographic Naive Targeting Under this scheme, the exact same transfer is given to all persons in the poorest areas (measured by the squared poverty gap), until the budget is exhausted. As a result, when a budget of 5 million somoni is distributed, the squared poverty gap decreases to 5.61 percent (compared to 5.57 percent under the optimal scheme). Approximately 34 percent of the poor would not benefit from this transfer, and 42 percent of the people receiving transfers are non-poor. In the case of a budget of 10 million somoni, the squared poverty gap decreases by 0.37 percentage points. 121 Table A12.2: Errors of Exclusion and Inclusion in Percentages (PMT package of 5 million somoni) Error of Error of Inclusion Exclusion Entire Population PMT 49.2 49.3 Geographic optimal 31.2 61.9 Geographic simple 42.4 34.1 Geographic optimal + PMT 32.9 83.6 Geographic simple + PMT 46.2 66.5 Pensioners PMT 43.2 51.5 Geographic optimal + PMT 29.3 82.3 Geographic simple + PMT 39.2 67.0 Proxy Means Testing The idea of a means-testing model is to approximate the welfare level of households using observable characteristics. The welfare level is commonly measured using a per capita (or per adult equivalent) expenditure aggregate that includes food expenditures, rent (paid or imputed), and the flow of services by durable goods, among others. The reason to use expenditures instead of incomes is partly justified by the low quality of reported income in surveys, and the belief that expenditure, by not being as volatile as income, more accurately reflects long-term welfare levels. The dependent variable for the means-testing regression models is the natural logarithm of the per capita expenditure aggregate, and the explanatory variables are selected from a set of variables assumed to be associated with welfare, such as geographic dummies, ownership of durable goods, characteristics of the dwelling, and gender and age of the head of household, among others. We search for models with good statistical fit and estimate the model using household survey data. The impact on poverty of transferring the same budget to the people identified as poor by the regression model83 is lower than for the geographic schemes (see Table A12.1). The first row in Table A12.2 presents the targeting performance of the model. The errors of inclusion and exclusion of this model are almost identical at 49 percent, and the errors of inclusion are the highest of all models considered here. Proxy Means Testing plus Geographic Targeting Looking for a way to potentially improve the targeting performance of transfers, it is possible to use the proxy-means-testing model in the areas selected by geographic targeting. When combining PMT with optimal geographic targeting, transfers are paid to households identified as poor by the PMT model in the regions selected by the optimal scheme. We re-compute the per capita transfers used in the geographic simulations so that the total amount paid to the predicted poor by PMT in any region matches what was paid under the optimal scheme. Both the errors of exclusion and inclusion are higher when compared to the geographic scheme without PMT. The changes in poverty84 are also lower than in the case of the optimal scheme. There seems not to be a good reason to mix these two schemes in Tajikistan, when judged by impact on poverty and targeting performance. 83 The threshold for poverty prediction was set such that the headcount ratio of predicted poor matches that ratio before transfers. The reason for doing this is that only 29 percent of the population would be predicted poor by PMT if we use the original poverty line as the threshold. 84 According to the squared poverty gap. 122 When combining PMT with the simple geographic targeting, transfers are paid to households identified as poor by the PMT model in the regions selected by the simple scheme. We re-compute the per capita transfers used in the geographic simulations so that the total amount paid to the predicted poor by PMT in any region matches what was paid under the simple scheme. As was the case with the optimal scheme, the poverty impact is lower and the errors of inclusion and exclusion are higher when combining the simple geographic scheme with PMT than the simple scheme alone. We also tried the five schemes described above (optimal, simple, PMT, optimal + PMT, simple + PMT), restricting the analysis to the old-age and disability pensioners. In the first case (Table A12.1, Panel B), we assume that the funds to be transferred under the different schemes came from an outside source, and in the last case (Table A12.1, Panel C), we assume that the funds come from uniformly reducing the old- age and disability pensions by a fixed percentage (20 percent in the case of a 5 million somoni budget and 40 percent for a 10 million somoni budget). This last scenario is called balanced budget for obvious reasons. The results are qualitatively the same as for the general population: there is no evidence in the data to support the use of a proxy-means-testing model to improve targeting performance and poverty reduction when compared to simpler geographic schemes. 123 Annex 13: Additional Tables Describing Education Sector Table A13.1: Average Expenditure per Student per Year, 2007 Somoni per Student Percentage Public Private Total Public Private Total Preprimary School 161 General Education 141 146 287 49.1% 50.9% 100 % Secondary Professional Education 348 345 693 50.2% 49.8% 100 % Higher Professional Education 142 836 978 14.5% 85.5% 100 % Source: World Bank 2008b, TLSS and authors' calculations. Table A13.2: Mean Private Expenditure on Public Schools per Year (Somoni) Text- books Meals Build- School School Educa- Other and other and/or ing Fees and Uni- tional Ex- Total Instruc- Lodg- Repair, Tuition forms Supplies penses tion ing etc. Materials All 36 90 16 17 23 8 5 191 Dushanbe 72 101 18 22 26 10 7 252 Sogd 40 78 17 18 31 14 5 193 KHatlon 24 85 15 13 15 4 2 163 RRS 33 99 15 20 25 8 8 211 GBAO 28 89 18 20 11 3 3 175 Dushanbe 72 101 18 22 26 10 7 252 Other urban 55 92 21 20 26 8 6 221 Rural 25 85 14 16 21 8 4 176 General Education 10 85 15 16 7 8 4 146 Secondary Special Education 119 98 22 20 114 6 10 345 Higher Education 396 128 33 28 226 12 22 836 Poorest 7 69 12 15 8 5 3 122 Poor 14 77 14 16 10 5 3 141 Middle 20 82 14 14 10 4 2 151 Rich 34 92 17 17 25 5 4 195 Richest 91 114 21 22 54 19 11 325 Source: TLSS 2007 and authors' calculations. 124 Table A13.3: Enrolment Rates at Ages 8­18, 2007 Region Ethnicity All Dushanbe Sogd Khatlon RRS GBAO Tajik Uzbek Others All 0.92 0.91 0.92 0.89 0.99 0.91 0.91 0.99 0.91 Male 0.98 0.92 0.97 0.94 0.99 0.96 0.93 1 0.95 Female 0.86 0.89 0.87 0.84 0.99 0.86 0.89 0.98 0.87 Age 8 9 10 11 12 13 14 15 16 17 18 All 0.99 0.99 0.99 0.99 0.98 0.96 0.97 0.94 0.81 0.66 0.37 Male 0.99 1 1 1 0.98 0.98 0.98 0.97 0.89 0.81 0.57 Female 0.99 0.99 0.98 0.99 0.98 0.93 0.96 0.9 0.74 0.51 0.24 Consumption Quintiles Poorest Poor Middle Rich Richest All 0.88 0.91 0.93 0.91 0.93 Male 0.93 0.95 0.96 0.96 0.97 Female 0.85 0.87 0.89 0.86 0.89 Source: TLSS 2007 and authors' calculations. 125 References Asad, Alam, Mamta Murthi, and Ruslan Yemtsov. 2005. "Growth, Poverty, and Inequality: Eastern Europe and the Former Soviet Union." Washington, D.C.: World Bank. ADB (Asian Development Bank). "Tajikistan Hospital Service Restructuring Concept for 2006­2010." Consultancy to the MOH financed by ADB. Dhusanbe. Dabalen, A., and W. Wane. 2008. "Informal Payments and Moonlighting in Tajikistan's Health Sector." Policy Research Working Paper 4555. World Bank, Washington, D.C., March. Deaton, A., and S. Zaidi. 1999. "Guidelines for Constructing Consumption Aggregates for Welfare Analysis." Princeton University Woodrow Wilson School, Development Studies Paper 192. Princeton, NJ. European Observatory on Health Systems. 2000. "Tajikistan Health System in Transition." Brussels, Belgium. Ivakhnyuk, I. 2006. "Migration in the CIS Region: Common Problems and Mutual Benefits." International Symposium on International Migration and Development, Population Division. Turin: Department of Economic and Social Affairs, United Nations Secretariat. Johnstone, Bruce. 2000. "Student Loans in International Comparative Perspective: Promises and Failures, Myths and Partial Truths." University at Buffalo Center for Comparative and Global Studies in Education: International Comparative Higher Education Finance and Accessibility Project (Summer 2003). ______, Bruce. 2004. "The Applicability of Income Contingent Loans in Developing and Transitional Countries." Journal of Educational Planning and Administration (New Delhi). Kanbur, Ravi. 1987. "Measurement and Alleviation of Poverty." International Monetary Fund Staff Papers. Washington D.C. MOE (Ministry of Education). 2005. "National Strategy for Education Development of the Republic of Tajikistan (2006­2015)." Dushanbe. MOH (Ministry of Health). 2007. "Assessment of Financial Burden and Degree of Satisfaction of Health Service Users at Level of PHC and Hospital Service Prior to Introduction of the BBP." Dushanbe. Nguyen, Trang V., and Ramya Sundaram. 2009. "Service Delivery in Europe and Central Asia: Recent Improvements and Remaining Challenges." In progress. Nguyen, Trang V., Ramya Sundaram, and Kathy Lindert. 2009. "The Redistributive Impact of Social Protection Systems in Europe and Central Asia." In progress. World Bank, Washington, D.C. OECD (Organisation for Economic Co-operation and Development). 2007. "Education at a Glance." Paris. Ravallion, Martin. 1994. Poverty Comparisons. Switzerland: Harwood Academic Press. TLSS (Tajikistan Living Standards Survey). 2007. The World Bank, United Nations Children's Fund (UNICEF) and Tajikistan State Statistics Committee. Dushanbe. Tajikistan State Committee on Statistics. 2006. Tajikistan Multiple Indicator Cluster Survey (MICS) 2005. Dushanbe. Tishkov, V., Zh. Zayinchkovskaya, and G. Vitkovskya. 2005. "Migration in the Countries of the Former Soviet Union." Report for the Global Commission on International Migration. Geneva. 126 UN (United Nations). 2007. "Tajikistan Situation Update 2007." Dushanbe. UNAIDS. 2008. "Eastern Europe and Central Asia: AIDS Epidemic Update." Regional Summary. Geneva. UNDP (United Nations Development Programme). 2005. "Investing in Sustainable Development: Millennium Development Goals Needs Assessment in Tajikistan." Dushanbe, Tajikistan UNICEF (United Nations Children's Fund). 2007. "Child Poverty in Tajikistan." New York. WFP/FAO (United Nations World Food Programme/Food and Agriculture Organization). 2008. "Joint Food Security Assessment." Dushanbe. World Bank. 2005. "Millennium Development Goals: Progress and Prospects in the Europe and Central Asia Region." Washington, D.C. ______. 2007. "Programmatic Public Expenditure Review. Health Sector Public Expenditure Tracking Survey." Washington, D.C. ______. 2008. "Health Sector Fiduciary Capacity Assessment Report." Washington, D.C. ______. 2008. "Welfare Dynamics and State of Service Delivery in the Health and Education Sectors." Draft. Washington, D.C. ______. 2008. "Welfare Dynamics and State of Service Delivery in the Health and Education Sectors." Washington, D.C. ______. 2008b. "Tajikistan Second Programmatic Public Expenditure Review." Report No. 43280-TJ. Washington, D.C. ______. Forthcoming. "Multisectoral Constraints to the Achievement of Health Outcomes Assessment." Washington, D.C. World Development Indicators. Various years. World Bank, Washington, D.C. 127